Understanding and producing the reduced...

Journal of

www.elsevier.com/locate/jml

Journal of Memory and Language 56 (2007) 410–435

Memory andLanguage

Understanding and producing the reduced relativeconstruction: Evidence from ratings, editing and corpora q

Mary Hare a,*, Michael K. Tanenhaus b, Ken McRae c

a Department of Psychology, Bowling Green State University, Bowling Green, OH, USAb University of Rochester, Rochester, NY, USA

c University of Western Ontario, London, Ont., Canada

Received 23 January 2006; revision received 8 August 2006Available online 4 October 2006

Abstract

Two rating studies demonstrate that English speakers willingly produce reduced relatives with internal cause verbs (e.g.,Whisky fermented in oak barrels can have a woody taste), and judge their acceptability based on factors known to influenceambiguity resolution, rather than on the internal/external cause distinction. Regression analyses demonstrate that fre-quency of passive usage predicts reduced relative frequency in corpora, but internal/external cause status does not. Theauthors conclude that reduced relatives with internal cause verbs are rare because few of these verbs occur in the passive.This contrasts with the claim in McKoon and Ratcliff (McKoon, G., & Ratcliff, R. (2003). Meaning through syntax: Lan-guage comprehension and the reduced relative clause construction. Psychological Review, 110, 490–525) that reduced rel-atives like The horse raced past the barn fell are rare and, when they occur, incomprehensible, because the meaning of thereduced relative construction prohibits the use of a verb with an internal cause event template.� 2006 Elsevier Inc. All rights reserved.

Keywords: Lexical semantics; Language comprehension; Language production; Meaning through syntax (MTS)

Introduction

More than three decades ago, Bever (1970) usedreduced relatives such as the horse raced past the barn fell

as the centerpiece of an argument that merely adding

0749-596X/$ - see front matter � 2006 Elsevier Inc. All rights reserv

doi:10.1016/j.jml.2006.08.007

q This work was supported by NIH grant MH6051701 to thefirst and third author, NIH grant HD27206 to the secondauthor, and NSERC grant OGP0155704 to the third author.We would like to thank Dana Subik for assistance withmaterials and analysis for Study 2, Doug Roland for his inputto Study 3, and Wendy Fogo for helping with the itemclassification in that study.

* Corresponding author. Fax: +1 419 372 6013.E-mail address: [email protected] (M. Hare).

performance considerations such as limited workingmemory capacity to a linguistic competence grammarwould be insufficient to explain linguistic performancein sentence perception. Bever argued that the field ofpsycholinguistics would have to develop its own inde-pendently motivated theories of how grammatical rela-tions are recovered during comprehension. The forceof Bever’s argument was that neither grammatical com-petence, nor memory limitations, nor any simple combi-nation of the two, would provide a plausible account forwhy the horse raced past the barn fell is typically judgedto be unacceptable. A grammatical explanation failsbecause propositions that can be realized in unreducedrelative clauses invariably can occur in reduced relativeclauses. There is no principled linguistic argument for

ed.

mailto:[email protected]

M. Hare et al. / Journal of Memory and Language 56 (2007) 410–435 411

why there should be an exception to this general charac-teristic of English grammar for the horse that was raced

past the barn fell. A memory load explanation failsbecause other sentences that seem to make heavierdemands on a limited capacity memory are clearlyacceptable, as for example, The horse that was presented

to the visiting dignitary as a gift by the Arabian Horse

Foundation was a truly magnificent beast.Bever’s alternative explanation for the difficulty of

the horse raced past the barn fell was based on his obser-vation that the horse raced past the barn is ambiguousbetween a reduced relative and a main clause. Bever(1970) proposed that readers and listeners are garden-pathed because they follow a general Subject-Mainverb-Object processing strategy, which biases themtowards the more canonical main clause and away fromthe less likely reduced relative. We should note that Bev-er’s perceptual strategies approach was a precursor ofrecent evidence-based approaches to ambiguity resolu-tion, all of which adopt essentially Bayesian perspec-tives. In the subsequent decades, especially the 1980sand 1990s, temporary ambiguity became a centralempirical testing ground for evaluating models of pars-ing and interpretation (for reviews, see Frazier, 1987;MacDonald, Pearlmutter, & Seidenberg, 1994; Tanen-haus & Grodner, 2006; Tanenhaus & Trueswell, 1995).Reduced relatives have featured prominently in manytheoretical and empirical debates, especially debatesabout whether certain classes of constraints, such asverb-specific information and referential context, influ-ence initial syntactic commitments, as claimed by con-straint-based models (MacDonald, 1993; McRae,Spivey-Knowlton, & Tanenhaus, 1998; Pearlmutter &MacDonald, 1995; Trueswell, 1996; Trueswell & Tanen-haus, 1991; Trueswell, Tanenhaus, & Garnsey, 1994), orthe ease of garden-path recovery, as claimed by two-stage models (Clifton et al., 2003; Ferreira and Clifton,1986; Frazier, 1995; Rayner, Carlson, & Frazier, 1983;Rayner, Garrod, & Perfetti, 1992).

Stevenson and Merlo (1997) observed that many diffi-cult reduced relatives contain unergative (manner-ofmotion) verbs such as race, whereas easier reduced rela-tives contain unaccusative (external cause change of state)verbs. Adopting the syntax-in-the-lexicon approach ofHale and Keyser (1993), they proposed that garden-pathrecovery for reduced relatives with unergative verbsexceeds the resource limitations of the competitive-at-tachment parser proposed by Stevenson (1994). McKoonand Ratcliff (2003, 2005) build upon Stevenson and Mer-lo’s observations about manner of motion verbs, but pres-ent a dramatically different approach. McKoon andRatcliff argue that there is a principled grammatical rea-son why the horse raced past the barn fell is unacceptable.They claim that verbs like race, which have an internalcause in their event template, are grammaticality prohib-ited from being used in the reduced relative construction

because internal control by the entity in head position isprohibited by the construction meaning of the reduced rel-ative. Thus, in contrast to the standard view in the field,they argue that the horse raced past the barn fell is unac-ceptable because it is not grammatically licensed. Thisanalysis of the reduced relative construction forms thecenterpiece of a new theory of sentence comprehensionand production, Meaning through Syntax (MTS). Thedata offered in support of the MTS analysis, presentedin McKoon and Ratcliff (2003, 2005), come from ratingstudies, reading time experiments, and most crucially, alarge-scale corpus analysis, from which the authors con-clude that reduced relatives occur so rarely with internalcause change of state verbs, such as erode, and mannerof motion verbs, such as race, that the few occurrencescan be considered errors.

McRae, Hare, and Tanenhaus (2005) critiqued theMTS approach to reduced relatives and defended ambi-guity-resolution accounts, in particular constraint-basedapproaches, which had been criticized by McKoon andRatcliff (2003). Here we build on our earlier work by pre-senting data from a rating study (which was briefly sum-marized in McRae et al., 2005) an editing study and acorpus analysis. The results are incompatible with theMTS approach, while supporting alternatives based uponmore standard linguistic and processing assumptions,including those incorporated into most constraint-basedmodels. Before we turn to the experimental work, wereview some of linguistic background that McKoon andRatcliff (2003) use to motivate MTS and its approach tothe reduced relative construction.

Verb meaning and verb argument structure

It is well known that there is a systematic relationshipamong the type of event or activity that a verb denotes,the nature of the entities that participate in that event,and the types of syntactic complements with which thatverb can occur. As one example, events described byverbs of transfer typically involve three participants:The agent of the transfer, the recipient, and the entitybeing transferred, which is the theme or patient. [Notethat although we adopt the terminology of thematicroles to describe modes of participation, the point weare making is independent of one’s theoretical stanceabout the usefulness of thematic roles as a linguistic orconceptual construct.] These verbs typically occur inthe dative construction (i.e., Noun Phrase Verb NounPhrase to Noun Phrase; Cole sent the flyer to the man)or the double object construction (Noun Phrase VerbNoun Phrase Noun Phrase; Cole sent the man the flyer),where the structural arguments in the syntax align withthe semantic arguments of the verb. Similarly, certainclasses of verbs denote events or activities with onlyone participant (She laughed; I sneezed) and these occurpredominantly in intransitive structures. Children as

412 M. Hare et al. / Journal of Memory and Language 56 (2007) 410–435

well as adult comprehenders exploit these correlationsbetween a verb’s meaning and its preferred argumentstructure (Fisher, Gleitman, & Gleitman, 1991; Gillette,Gleitman, Gleitman, & Lederer, 1999; Hare, McRae, &Elman, 2004).

Observations like these underlie both the syntacticbootstrapping hypothesis for how young childrenmight learn words whose meaning cannot easily bedepicted (Gleitman, 1990) and the competing semanticbootstrapping hypothesis, which argues that the childrelies on the cognitive distinction between things andactions to develop the syntactic categories of nounand verb (Pinker, 1989). The non-arbitrary mappingbetween syntactic complements and verb meaning alsoprovides some of the motivation for the influentialresearch program of Levin and colleagues (Levin,1993; Levin & Rappaport Hovav, 1995), which usesshared alternations like the to-dative and double-objectconstruction to construct more fine-grained verb clas-ses. This work, along with the notion of constructionmeaning proposed in Goldberg (1995), provides thefoundation for MTS, and so we review the most rele-vant findings before turning to the details of theMTS account of reduced relatives.

Levin and Rappaport Hovav (1995), among others,have noted the close fit between syntactic structure andthe causal structure of the event denoted by the verb.As a general rule, the entity construed as the cause ofan event or activity appears in subject position, whilean entity that changes as a result of the event, or other-wise undergoes its effects, appears as direct object. Thisintuition is formalized as an event template that repre-sents those aspects of a verb’s meaning (including cau-sality) that determine the syntactic structures in whichthe verb will occur. On Levin and Rappaport Hovav’saccount, there are two types of causal structure, external

and internal. External cause verbs such as break havetwo arguments, the entity that becomes broken and anagent, instrument, or natural force that causes thebreaking event. In contrast, internal cause verbs suchas walk or erode have one argument, which is responsi-ble for bringing about the activity or event in which itparticipates. Linking or movement rules relate the causalargument in both cases to syntactic subject, and (forexternal cause verbs) link the affected argument to directobject position. As a result, external cause verbs likebreak are predicted to occur in transitive sentences (Ibroke the dish), and internal cause verbs in intransitives(The dog walked). These are the dominant syntactic pat-terns for these verbs, but other structures are alsoacceptable, and therefore this account includes mecha-nisms to account for transitivity alternations. Theseallow an external cause argument not to surface withexternal cause verbs (The dish broke) or be added tothe syntax with internal cause verbs (High waves eroded

the beach; I walked my dog along the embankment).

The apparently straightforward relationship betweenthe syntax and lexical semantics is complicated by theimportant observation that syntactic constructions them-selves can contribute to meaning (Goldberg, 1995, 2003;Goldberg & Jackendoff, 2004; Jackendoff, 1997, 2002a,2002b). Some of the strongest evidence comes from theobservation that constructions can coerce interpretationsthat are clearly not licensed by the verb itself. For exam-ple, the caused-motion construction, exemplified by sen-tences such as Ken finally hit the ball out of the infield,expresses the meaning that the cause argument (Ken)causes the theme (the ball) to move along the path indicat-ed by the directional phrase (out of the infield), and thesentence is interpreted that way even though hit is not acaused motion verb. More strikingly, the same interpreta-tion is available in Fred sneezed the tissue off the table, orThey laughed the poor guy out of the room, which aregrammatical even though neither sneeze nor laugh licens-es a direct object. The caused-motion meaning is contrib-uted by the construction itself, and the verb is said to becoerced into that meaning by the construction.

Meaning through syntax

With MTS, McKoon and Ratcliff (2003) adopt ideasabout event templates taken from the work of Levin andcolleagues, and ideas about construction meaning thatare inspired by construction grammar, and combinethese in a novel way to derive empirical predictions aboutsentence processing, in particular comprehension of thereduced relative construction. On the MTS account,the reduced relative construction expresses ‘‘participa-tion in an event caused by some force or entity externalto itself’’ (McKoon & Ratcliff, 2003, p. 492). Internalcause verbs cannot appear in reduced relatives in whichthe head is the entity engaging in the activity because thisis inconsistent with the meaning of the construction.Coercion (in the standard construction grammar sense)does not play a role, and therefore reduced relatives likeThe horse raced past the barn fell simply violate thisrestriction and are deemed ungrammatical. In contrast,external cause change of state verbs like break or fade

abide by the restriction, and are permitted in the reducedrelative (McKoon & Ratcliff, 2003, p. 496). Thus MTSclaims that a simple dichotomous variable, presence orabsence of an external cause in a verb’s event template,coupled with the meaning of the reduced relative con-struction, can account for whether or not a verb islicensed to participate in a reduced relative.

This claim is strong and counter-intuitive, and thusdeserves careful scrutiny. It also leads to three clearempirical predictions. First, the status of a verb as exter-nal or internal cause will determine its acceptability andlikelihood of occurrence in reduced relative construc-tions. Reduced relatives with external cause verbs willbe acceptable (McKoon & Ratcliff, 2003, p. 496), while


those in which the head is an internal cause will not be(McKoon & Ratcliff, 2003, p. 502). Consequently,MTS also predicts that reduced relative comprehensionwill be unrelated to ambiguity, contrary to the more gen-erally accepted prediction that pragmatic, discourse andsemantic factors that are known to influence ambiguityresolution should affect the acceptability of sentenceswith reduced relatives. Finally, MTS states that differentsyntactic forms interact differently with the meaning ofverbs (McKoon & Ratcliff, 2005, p. 1035), such thatthe ungrammaticality of a verb in the reduced relativeis unrelated to its acceptability in the unreduced relativeor the passive. Again, this contrasts with the standardaccount, assumed by linguistic analyses and ambiguity-based accounts, that the reduced relative is a type of pas-sive construction, and the degree to which any particularverb occurs in the simple passive should predict thatverb’s likelihood of occurring in other passive construc-tions, like the reduced relative.

We present two questionnaire studies and a corpusanalysis that were designed to contrast these predictionsof the MTS event template approach with those made bystandard ambiguity-based accounts, including both con-straint-based accounts and two-stage models (Cliftonet al., 2003; Ferreira & Clifton, 1986; MacDonaldet al., 1994; McRae et al., 1998; Rayner et al., 1983; Sturt,Scheepers, & Pickering, 2002; Tanenhaus, Spivey-Knowlton, & Hanna, 2000; Tanenhaus & Trueswell,1995). In the first questionnaire study, participants ratedthe comprehensibility of sentences with unreduced rela-tive clauses, reduced relative clauses, and passives for setsof (a) external cause change of state verbs and (b) inter-nal cause change of state and manner of motion verbs.For each set, we constructed reduced relatives that weredesigned to be easy or hard based on factors that areknown to influence ambiguity resolution. The resultsdemonstrate that there is no categorical distinction inthe comprehensibility of reduced relatives with internalor external cause verbs, that factors underlying ambigu-ity resolution determine comprehensibility, and that thethree passive structures are related. In the second ques-tionnaire study, we asked participants to produce short-ened versions of target sentences, the most important ofwhich were the unreduced relatives from Study 1. Easyreduced relatives for internal cause verbs were frequentlygenerated, and received high acceptability ratings, andthere was a strong relationship between reduced andunreduced relatives. These studies led us to conduct alarge-scale corpus study using the two most widely vettedparsed corpora, Brown and the Wall Street Journal, andtwo larger corpora, the Wall Street Journal 87 and theBritish National Corpus. The goal was to determine ifverb template, i.e., whether or not a verb is hypothesizedto have an external cause, explains the occurrence ofreduced relatives, when a less construction-specific fac-tor, frequency of passive use, is taken into account.

Study 1

Participants were presented with a set of sentences, asdescribed above, and were asked to rate how easy eachsentence was to understand on a seven-point scale, where1 = ‘‘makes no sense’’, 4 = ‘‘moderately easy to under-stand’’, and 7 = ‘‘extremely easy to understand’’. [Partic-ipants were actually given the reverse of this scale, but wereverse it here for presentation purposes so that it match-es the scale used in Study 2. This conversion of responsesdoes not influence the inferential statistics in any way.]Study 1 has three main goals. The first is to investigatewhether there is a categorical distinction between exter-nal and internal cause verbs in their potential usage inreduced relatives, contradicting the MTS position thatreduced relatives with an internal cause in head positionare prohibited in English. If MTS was correct, internalcause reduced relatives would not be rated as easily com-prehensible. Thus our first goal is to test whether reducedrelatives with internal cause verbs can be easily compre-hended. The second goal is to test whether, contrary tothe assertions of MTS, difficulty in comprehendingreduced relatives is related to their temporary ambiguity.To investigate this, we compare ratings of unreduced(unambiguous) and reduced (ambiguous) relative claus-es. We also argue that the ability of a verb to appear inthe reduced or unreduced relative is dependent on itsability to appear in the passive, because all three are pas-sive constructions. Therefore, the final goal is to testwhether the acceptability of a verb in the reduced andunreduced relatives is related to acceptability of the cor-responding passive, and that acceptability of the passiveis a better predictor of reduced relative acceptability thanis the internal/external cause distinction.

To accomplish these three goals, we chose 24 externalcause and 23 internal cause verbs, and created a sentencewith a reduced relative clause that was predicted to behard to understand for 12 of the verbs of each type,and a sentence that was predicted to be easy to under-stand for the other 12 (with one internal cause verb usedtwice). Using the reduced relative as the kernel sentence,we then transformed it into a sentence with an unreducedrelative clause and one with a simple passive, and askedparticipants to rate the comprehensibility of each sen-tence type. The easy/hard difference was based on thespecific verb, plus factors that have been identified asmodulating the difficulty of reduced relative clauses.

Method

Participants

Thirty-three members of the Psychology communi-ties at Bowling Green State University and the Universi-ty of Western Ontario received course credit or werepaid for their participation. There were 11 participantsin list 1, 10 in list 2, and 12 in list 3.


Materials

We constructed sentences using 24 external causeand 23 internal cause verbs (the internal cause verbwalked was used in both the easy and hard internalcause conditions). Causal status of the verbs was deter-mined by McKoon and colleagues’ classifications(McKoon & MacFarland, 2000, 2002; McKoon & Rat-cliff, 2003). All verbs had identical past tense and pastparticiple forms. For each verb, reduced relative, unre-duced relative, and passive sentences were constructed.In addition to manipulating the verb, we created easyand hard sentences. The hard sentences were designedto be difficult to interpret as a passive construction,while the easy sentences were designed to support thatinterpretation. To do this, we manipulated the thematicfit of the initial noun phrase (the subject of the passiveor the head of the unreduced or reduced relative), usingnoun phrases that were more likely to be agents in theevent denoted by the verb in the hard sentences (e.g.,waiter served), and noun phrases that were more likelyto be patients in the easy ones (e.g., applicants inter-

viewed). In addition, we manipulated the presence andtype of post-verbal prepositional phrases to increaseor decrease difficulty. The easy items with externalcause verbs contained good patients in all 12 cases,postverbal agentive by-phrases in 7 cases, and variousprepositional phrases or a postverbal verb phrase inthe other 5 cases. In contrast, the hard items withexternal cause verbs contained good agents in all 12cases, a good patient directly following the verb in 6cases (no preposition), and various prepositional phras-es or a postverbal verb phrase in the other 6. The easyitems with internal cause verbs contained good patientsin all 12 cases, postverbal agentive by-phrases in 8cases, various prepositional phrases in 3 cases, and anadverb in the remaining case. The hard items withinternal cause verbs contained good agents and tempo-ral or locative prepositional phrases in all 12 cases,with no by-phrases. The motivation for these factorscomes from empirical results reported in Clifton et al.(2003), Ferreira and Clifton (1986), MacDonald(1993), McRae et al. (1998), Rayner et al. (1983), True-swell et al. (1994), Spivey-Knowlton, Trueswell, andTanenhaus (1993). All items are presented with theirmean ratings in Appendix A.

Table 1Sentence comprehensibility ratings by condition for Study 1

Construction External cause

Easy Hard

M SE M

Reduced relative 6.0 0.1 2.9Unreduced relative 6.1 0.2 5.6Passive 6.5 0.1 6.1

We constructed three lists, each containing 4 itemsfrom each of the 12 conditions that were formed bycrossing verb type (external vs. internal cause), difficulty(easy vs. hard), and construction (reduced relative, unre-duced relative, passive). Thus each participant rated 16reduced relatives, 16 unreduced relatives, and 16 pas-sives. Each list also contained 96 filler sentences of var-ious other constructions, lengths, and levels of difficulty.

Procedure

Participants were instructed that 144 sentences wouldbe presented, one at a time. They were asked to rate howeasy each sentence was to understand on a 7-point scalewith 1 = ‘‘makes no sense’’, 4 = ‘‘moderately easy tounderstand’’, and 7 = ‘‘extremely easy to understand’’.They were asked to try to use the entire scale in their rat-ings. Participants also were informed that judgment timewas not of interest, so they were free to work at theirown pace. Sentences were presented one at a time in ran-dom order on a color monitor connected to a Macintoshcomputer, with stimulus presentation and data record-ing controlled by PsyScope (Cohen, MacWhinney, Flatt,& Provost, 1993). The response scale remained on thescreen below the sentence during each trial. Participantsrecorded their ratings by pressing the 1 through 7 keys atthe top of the keyboard.

Results and discussion

Mean sentence comprehensibility ratings for all con-ditions are presented in Table 1, and the analyses of var-iance results are presented in Table 2.

Reduced relative comprehensibility

The dependent variable was reduced relative compre-hensibility rating, and verb type (external cause vs. inter-nal cause) and difficulty (easy vs. hard) were theindependent variables. Both independent variables werewithin participants, but between items. In all relevantanalyses reported in this article, list was included as abetween-participants dummy variable and item rotationgroup as a between-items dummy variable to stabilizevariance that may result from rotating participantsand items over lists (Pollatsek & Well, 1995). Effectsinvolving these dummy variables are not reported.

Internal cause

Easy Hard

SE M SE M SE

0.2 5.6 0.2 3.3 0.20.2 6.1 0.1 4.8 0.20.1 6.4 0.1 5.1 0.2

Table 2Analyses of variance for Study 1

Effect Participants Items minF 0

Reduced relative clauses only

Difficulty F(1,30) = 321.78 F(1,36) = 409.80 F(1,63) = 180.25Verb type F < 1 F < 1 F < 1Interaction F(1,30) = 11.14 F(1,36) = 11.04 F(1,66) = 5.54Anomalous Fillers vs. Hard Internal F(1,32) = 38.69 F(1,17) = 9.24 F(1,25) = 7.46Easy Fillers vs. Easy Internal F < 1 F < 1 F < 1

Ambiguity, difficulty, and verb type

Difficulty F(1,30) = 338.68 F(1,36) = 187.13 F(1,64) =120.53Ambiguity F(1,30) = 224.85 F(1,36) = 190.18 F(1,66) = 103.03Ambiguity · Difficulty F(1,30) = 83.44 F(1,36) = 111.36 F(1,63) = 47.70Verb type F(1,30) = 6.76 F(1,36) = 2.32ns F(1,57) = 1.73ns

Verb type · Ambiguity F(1,30) = 6.88 F(1,36) = 3.66ns F(1,63) = 2.39ns

Verb type · Difficulty F < 1 F < 1 F < 1Three-way interaction F(1,30) = 28.06 F(1,36) = 20.73 F(1,66) = 11.92

Note. all F’s significant unless otherwise noted. ns = non-significant.


In all studies reported here, the halfwidth of the confi-dence interval of the difference between means is report-ed in parentheses with the relevant contrasts. If thehalfwidth is less than the observed difference betweenmeans, then the contrast would be significant by a con-ventional inferential test. Confidence intervals werecomputed using the methods suggested by Masson andLoftus (2003).

Factors influencing ambiguity resolution had a largeeffect on comprehensibility ratings: Easy reduced rela-tives, M = 5.8, were rated as significantly more compre-hensible than hard ones, M = 3.1 (0.3). Crucially,comprehensibility ratings were almost identical forexternal cause, M = 4.4, and internal cause verbs,M = 4.4 (0.3). Thus verb type did not influence compre-hensibility ratings, in contrast to the MTS claim thatreduced relatives are comprehensible with externalcause, but prohibited with internal cause verbs.

In addition, verb type and difficulty interacted. Easysentences with internal cause verbs were more compre-hensible than hard sentences with external cause verbs(mean difference = 2.7, halfwidth = 0.38). This result isinexplicable on the MTS account: If internal causereduced relatives are prohibited, they should not be rat-ed as more comprehensible than the allowable reducedrelatives with external cause verbs. Furthermore,although easy reduced relatives with external causeverbs were rated as more comprehensible than easyreduced relatives with internal cause verbs (mean differ-ence = 0.4, halfwidth = 0.38), hard reduced relativeswith internal cause verbs were rated as more comprehen-sible than hard reduced relatives with external causeverbs (mean difference = 0.4, halfwidth = 0.38).

Two further tests were conducted to provide bench-marks for the ratings for hard and easy reduced relativeclauses with internal cause verbs. In the first, we con-

trasted the ratings for the internal cause verbs with sevenfiller sentences. These fillers were all grammaticallyacceptable, but some were semantically anomalous,while others did have meaning but were difficult tounderstand (see Appendix B). These seven sentences(M = 2.2, SE = 0.1) were rated as significantly less com-prehensible than even the hard internal cause reducedrelatives (M = 3.3, SE = 0.2). More importantly, we alsocontrasted the ratings for the 12 easy internal cause sen-tences with those of 12 grammatical and perfectly com-prehensible filler sentences (also presented in AppendixB). Both groups had a mean rating of 5.6 out of 7, witha standard error of 0.2. Together, the two benchmarksdemonstrate that the high ratings for the internal causereduced relatives were not due to a participant strategyof accepting grammatically coherent strings whether ornot they were interpretable semantically. High ratingswere reserved for sentences that were both grammaticaland semantically coherent.

In summary, the results from the comprehension taskprovided no support for the hypothesis that reduced rel-atives with internal cause verbs are prohibited. Reducedrelatives with internal cause verbs were rated as highly asthose with external cause verbs, and in fact easy reducedrelatives with internal cause verbs were rated as signifi-cantly more comprehensible than the hard reduced rela-tives with external cause verbs, and equal togrammatical and semantically coherent sentences ofother types.

Because of considerations discussed later, in Study 3,where we consider precisely which reduced relatives aremost relevant to the claims of MTS, we eliminated threeitems with manner of motion verbs in which one mightargue that the head is not the participating entity. Theseare The path traveled by many settlers extended far to the

west, The mountain climbed by the tourists sloped gently


upward, and The city streets roamed by gangs of young

men were too dangerous for tourists. Removing theseitems did not affect the interpretation of any of the anal-yses because the mean comprehensibility rating for theeasy internal cause sentences changed only minimally,from 5.6 to 5.5, and any change in the inferential statis-tics was minimal as well, with no changes in the patternof effects.

Ambiguity effects

Our second goal was to show that much of the vari-ability in the comprehension of reduced relatives is relat-ed to temporary ambiguity. The first piece of evidencefor this was the main effect of difficulty described above.Further evidence comes from a comparison of the com-prehensibility ratings for unreduced and reduced rela-tives. If the difficulty is indeed due to ambiguity, thenthe ambiguous reduced relatives should be rated as lesscomprehensible than the unambiguous unreduced rela-tives. Furthermore, ambiguity (unreduced vs. reducedrelatives) should interact with difficulty (hard vs. easy),with larger ambiguity effects for hard than for easy sen-tences. This interaction should be found for both theexternal cause and internal cause verbs. Here we presentanalyses of variance testing these predictions.

We conducted analyses of variance in which verbtype was within participants but between items, ambigu-ity was within participants and within items, and diffi-culty was within participants but between items.Overall, easy sentences (M = 5.9) were rated as morecomprehensible than hard sentences (M = 4.1). Therewas a main effect of ambiguity, with unreduced sentenc-es (M = 5.6) receiving higher comprehensibility ratingsthan reduced sentences (M = 4.4). Collapsed across verbtype, ambiguity and difficulty interacted because therewas a larger effect of ambiguity for hard sentences (unre-duced: M = 5.2, reduced: M = 3.1) than for easy sen-tences (unreduced: M = 6.1, reduced: M = 5.8), aspredicted by two-stage and constraint-based ambiguityaccounts. The overall difference between internal(M = 4.9) and external (M = 5.1) cause verbs was signif-icant by participants, but not by items. Collapsed acrossdifficulty, the interaction between verb type and ambigu-ity was significant by participants, and marginal byitems. This occurred because the difference between theinternal cause unreduced (M = 5.5) and reduced sen-tences (M = 4.4) was smaller than for the external causeunreduced (M = 5.8) and reduced sentences (M = 4.4).Note that this difference is due solely to the unreducedsentences; comprehensibility ratings were equal for thereduced relatives. Verb type did not interact withdifficulty.

Finally, there was a three-way interaction amongverb type, difficulty, and ambiguity. Our main predictionfor these analyses was that difficulty and ambiguityshould interact for both internal and external cause

verbs. And indeed, although the three-way interactionresulted from a larger interaction for external cause thanfor internal cause verbs, both interaction contrast effectswere much larger than the relevant confidence interval.For the external cause verbs, the difficulty by ambiguitycontrast effect was 2.7, with a halfwidth of 0.4. For theinternal cause verbs, the contrast effect was 1.1, withthe same halfwidth. Both of these two-way simple inter-action effects occurred because the ambiguity effect waslarger for hard than for easy sentences, as predicted. Forinternal cause verbs, ratings were 0.5 higher for easyunreduced than for easy reduced sentences (0.5),whereas they were 1.5 higher for the unreduced hardsentences. For the external cause verbs, ratings wereonly 0.1 higher for easy unreduced than for easy reducedsentences, whereas they were 2.7 higher for the unre-duced hard sentences.

In summary, ambiguity interacted with difficultyoverall, and within each verb class. These interactionsoccurred because the ambiguity effects were substantial-ly larger for sentences designed to be hard than for thosedesigned to be easy.

Relations among constructions

On the MTS account, the meanings of the passive,unreduced relative, and reduced relative are different(cf. McKoon & Ratcliff, 2003, pp. 490, 502–503, 513).Importantly, the differences are assumed to influenceverb usage in the different constructions. Or, as McKoonand Ratcliff (2005) argue, ‘‘different syntactic formsinteract differently with the meanings of verbs. Theinteraction of race’s event template meaning with thesimple transitive structure or the simple passive structureleads to comprehensible sentences but the interactionwith the reduced relative structure does not.’’ (p. 1035,their italics). Note that the comment about race isintended to be true of all internal cause verbs.

We adopt instead the standard assumption that thesethree constructions are closely related in a way thatargues against the claim that they interact differentlywith the same verbs. All are types of passive construc-tions, and consequently in all cases the head or subjectnoun phrase must be interpreted as experiencing oraffected in some way by the action of the verb, ratherthan causing it. If this is true, then it should have conse-quences for the comprehensibility ratings. Our predic-tion was that ratings for reduced relatives, unreducedrelatives, and passives should be correlated, and thuspassive and unreduced relative comprehensibility ratingsshould predict reduced relative ratings. MTS predictsinstead that comprehensibility of the reduced relativeis predicted by verb type.

We used the 48 sentence triads (reduced relative,unreduced relative, and passive) created by crossing verbtype and difficulty. We calculated Pearson correlations(across the 48 items) between each pair of constructions


and found significant correlations (p < .001 in all cases):reduced relative and passive, r = .53; reduced and unre-duced relative, r = .56; and unreduced relative and pas-sive, r = .53. This demonstrates that comprehenders aresensitive to the relationships among these three types ofpassive constructions.

As a stronger test, we conducted a stepwise regres-sion analysis in which the dependent variable was thereduced relative comprehensibility rating, and the inde-pendent variables were the dichotomous internal/exter-nal cause distinction, the passive comprehensibilityrating, and the unreduced relative comprehensibility rat-ing. Unreduced relative rating entered first, followed bypassive comprehensibility rating, predicting 39% of thevariance in reduced relative comprehensibility ratings,F2(1,45) = 14.30 (for all inferential statistics reportedin this paper, p < .05 unless otherwise stated). Each ofthe two variables predicted significant unique propor-tions of variance (unreduced relative: r2 = .15,t2(45) = 2.85; passive: r2 = .11, t2(45) = 2.34). The inter-nal/external cause distinction did not enter (p = .94before any variables had entered, p = .30 after unre-duced relative had entered, p = .09 after both unreducedrelative and passive had entered).

In summary, Study 1 makes three important points.First, reduced relative clauses with internal cause verbsare not always hard to understand. Second, the resolu-tion of temporary structural ambiguity is an importantfactor in the comprehension of reduced relatives. Third,the passive, unreduced relative, and reduced relativeconstructions are related, as reflected in the related com-prehensibility judgments across the three constructions,and the fact that unreduced relative and passive compre-hensibility judgments predict the comprehensibility ofthe reduced relative. In Study 2, we test whether a sim-ilar pattern of results is found in a production task.

Study 2

We created all of the test sentences for Study 1.Therefore, it could be argued that whereas participantsmight judge some experimenter-generated reduced rela-tives with internal cause change of state and mannerof motion verbs to be acceptable, participants wouldnever willingly produce reduced relatives using theseverbs. To create a situation that would facilitate, butnot coerce, participants to produce sentences containingreduced relatives, we used a variation of a task takenfrom the Comprehensive Assessment of Spoken Lan-guage (Carrow-Woolfolk, 1999). In this task, grammat-ical proficiency in children is assessed by having anexperimenter read aloud sentences containing nounsmodified by prepositional phrases (e.g. The dog with long

hair jumped over the fence) and asking the child to short-en these into grammatical sentences with the same mean-

ing (e.g. The long-haired dog jumped over the fence). Inour version of the sentence-shortening task, we present-ed participants with 35 filler sentences intermixed withthe unreduced relative and the passive versions of thesentences used in Study 1, and asked them to generatea sentence similar in meaning to the original but at leasttwo words shorter. This procedure was designed to eval-uate the MTS hypotheses about the role of the internal/external cause distinction in the production and accept-ability of the reduced relative construction, and to allowus to look more closely at the relationship between thereduced and unreduced relative.

Analogous to Study 1, Study 2 had three major goals.The first was to test whether English speakers wouldproduce reduced relatives with internal cause verbs inan experimental setting, and, if they did, whether theywould rate their own productions as acceptable. The sec-ond goal was to test whether ambiguity influences theproduction of reduced relatives. If it does, then wewould expect participants to generate more reduced rel-atives from easy than from hard unreduced relatives,and to rate their own productions more acceptable forthe easy than for the hard items. Finally, we again testedwhether the constructions are related. This was accom-plished in two ways. First, participants were asked torate the similarity in meaning of the original unreducedrelatives and their shortened productions. In addition,we asked participants to rate acceptability of the originalunreduced relatives and passives. We conducted regres-sion analyses on the proportion of reduced relatives pro-duced for each item, with unreduced relativeacceptability, passive acceptability, and internal/externalcause verb status as the predictor variables.

Method

Participants

Twenty undergraduates from the University ofRochester were paid for their participation.

Materials

The 24 unreduced relatives and 24 matched passivesfrom Study 1 were used as test sentences. There were also35 filler sentences. Crucially, none of the sentences inStudy 2, whether test or filler item, contained a reducedrelative. This was done to avoid inflating the accessibilityof the reduced relative construction. Three of the fillersentences were intended to be semantically anomalous,as in Colorless green ideas sleep furiously, and 10 weredesigned to make it difficult to eliminate two words ormore and still create a grammatically acceptable sen-tence. These fillers were included to create some variationin the acceptability of the sentences that participantswould rate, and to provide some examples where it wasimpossible to create an acceptable sentence or a sentencewith similar meaning to the target sentence.


We created two lists by assigning the unreduced rela-tive and passive for each verb to a different list. The fill-ers and test sentences were then randomly intermixed,with the restriction that each half of the list containedthe same number of test sentences for each of the eightconditions created by crossing verb type (internal andexternal cause), type of construction (passive and unre-duced relative), and predicted reduced relative difficulty(easy and hard). We then created two random orders foreach list, resulting in four lists. Each participant wasassigned randomly to a single list.

Procedure

Participants were given a rating form with theinstructions and the sentences, and were instructed toread each sentence and rate its acceptability as an Eng-lish sentence on a scale of 1–7, where 1 was ‘‘extremelyunacceptable’’ and 7 was ‘‘completely acceptable’’. Theywere asked to shorten the sentence by deleting two ormore words, without adding words or rearranging thosealready present, and to write out the new sentence. Par-ticipants were asked to try to preserve the meaning ofthe original sentence, but were told that this would notalways be possible. They were also told that it mightnot always be the case that the shortened sentence wouldbe grammatically correct. After re-writing the sentence,they were asked to rate the acceptability of the shortenedversion on the same 7-point scale. In addition, they wereasked to rate how similar in meaning it was to the origi-nal on a 5-point scale, with 1 indicating ‘‘very differentmeanings’’ and 5 ‘‘the same meaning’’. Participants weregiven three examples (one of which could not be short-ened grammatically). The full instructions, the examplesentences, and the form of the response sheet are pre-sented in Appendix C.

Note that it was possible to shorten each target sen-tence in a number of ways. For example, the unreducedrelative sentence The guard who was searched inside the

prison walls had cocaine hidden in his jacket could beshortened into the reduced relative The guard searched

inside the prison walls had cocaine hidden in his jacket,but also into other constructions like The guard who

was searched inside the prison had cocaine in his jacket,The guard who searched inside the prison had cocaine in

his jacket, or The guard had cocaine in his jacket, amongother possibilities. In fact, all unreduced relatives couldbe shortened by deleting the entire relative clause. Thus,although participants could produce reduced relativeversions of the unreduced relative sentences, they werenot required to do so—other grammatical options wereavailable.

Design

Ninety-three percent of the shortened versions of theunreduced relative sentences were grammatical, asjudged by a research assistant and checked by the second

author. Analyses of variance were conducted on theseitems, using participants (F1) and items (F2) as randomvariables. Separate analyses were conducted for eachof three dependent variables: Proportion of reduced rel-atives produced, reduced relative acceptability ratings,and ratings of similarity between the original unreducedrelative and shortened reduced relative sentences. Theindependent variables were verb type (internal or exter-nal cause) and difficulty (easy or hard). Both variableswere within subjects and between items. The analysesof ratings of reduced relative acceptability, and ratingsof similarity between the original unreduced relativesand the shortened reduced relative were conducted byitems only. This was necessary because the independentvariables were within participants, and so if a partici-pant produced no reduced relatives in one or more ofthe four conditions, they would be excluded from theanalysis. As a result, only 7 of 20 participants wouldhave been included. In contrast, the independent vari-ables were between items, so this aspect of the analysiswas not a problem in the by-items analyses (at leastone participant produced a reduced relative for everyeasy external and easy internal cause item, and for 8of the hard external cause and 9 of the hard internalcause items). All confidence intervals are based on theitems analyses.

Correlation and regression analyses are reported aswell, with proportion of reduced relatives produced asthe dependent variable, and unreduced relative accept-ability, passive acceptability, and internal/external causeverb status as the predictors.

Finally, note that for all analyses in both Studies 2and 3 in which the dependent variable was a proportion,we also conducted the analyses using arc sine transfor-mations. Because the results were the same in every casein both studies, we report only the analyses based onuntransformed proportions.


Condition means for each dependent variable arepresented in Table 3, and the analyses of variance arepresented in Table 4.

Reduced relative probability and acceptability

According to MTS, participants should avoid gener-ating reduced relatives with internal cause verbs becausethey are prohibited. However, the data do not supportthis prediction. Instead, participants shortened the inter-nal cause unreduced relatives into reduced relatives witha probability of .34. The probability of reduced relativescreated with external cause verbs (.42) was marginallyhigher (halfwidth = 0.10). However, it is important tonote here that MTS predicts that reduced relatives withinternal cause verbs should be prohibited, not slightlydisfavored. Therefore we conducted a z-test of the differ-

Table 3A number of measures by condition for Study 2

Dependent variable External cause Internal cause

Easy Hard Easy Hard

M SE M SE M SE M SE

Proportion reduced relative 0.64 0.04 0.19 0.06 0.52 0.05 0.17 0.04Reduced relative acceptability 6.7 0.1 5.9 0.3 6.5 0.1 6.3 0.2Other acceptability 6.5 0.2 6.3 0.1 6.5 0.1 6.1 0.1Unreduced/reduced similarity 4.8 0.1 3.9 0.5 4.9 0.1 4.9 0.1Other similarity 4.1 0.1 3.6 0.2 4.3 0.1 3.9 0.1Unreduced relative acceptability 6.5 0.1 6.2 0.1 6.4 0.1 5.6 0.2Passive acceptability 6.8 0.1 6.5 0.1 6.6 0.1 5.8 0.2

Note. Acceptability was rated on a seven-point scale, whereas similarity was rated on a five-point scale. We used different scales foracceptability and similarity to reduce confusion for participants.

Table 4Analyses of variance for Study 2

Analysis Effect Participants Items minF 0

Production Verb type F(1,18) = 6.94 F(1,40) = 2.83ns F(1,58) = 2.01ns

Production Difficulty F(1,18) = 64.14 F(1,40) = 78.13 F(1,46) = 35.22Production Interaction F(1,18) = 2.59ns F(1,40) = 1.28ns F < 1Acceptability Verb type F < 1Acceptability Difficulty F(1,33) = 8.40Acceptability Interaction F(1,33) = 2.76ns

Similarity Verb type F(1,33) = 7.17Similarity Difficulty F(1,33) = 3.62ns

Similarity Interaction F(1,33) = 4.22

Note. all F’s significant unless otherwise noted; ns = non-significant.


ence between proportions to test whether the proportionof reduced relatives produced for internal cause verbswas significantly greater than 0. Because 0 cannot beused as the baseline proportion in this test (it wouldresult in dividing by 0), we used 0.001 as the MTS pre-dicted proportion of reduced relatives for internal causeverbs. The proportion of reduced relatives produced wassignificantly greater than .001 for both the easy internalcause items, .52, z = 56.90, and the hard internal causeitems, .17, z = 18.53.

The interaction between difficulty and verb type wasnon-significant because verb type had roughly the sameinfluence on reduced relative production for both levelsof difficulty. The proportion of produced reduced rela-tives did not differ for easy external versus internal causeitems (mean difference = 0.12, halfwidth = 0.13), nor forthe hard external versus internal cause items (mean dif-ference = 0.02, halfwidth = 0.13).

Because the unreduced relatives were taken fromStudy 1, they included the three easy internal cause verbitems containing path nominals. Although without theseitems the proportion of produced reduced relatives forthe easy internal cause items drops from 0.52 to 0.45,the main conclusions drawn from these data do notchange. That is, participants still produced reduced rel-

atives for 45% of the easy internal cause items, and aswas shown above, even the 17% produced reduced rela-tives for hard internal cause items was significantlygreater than 0.001, which was used as the baseline inthe z-tests.

The finding that over a third of the internal causeunreduced relatives were shortened into reduced rela-tives is clear evidence that reduced relative productionis not governed by a binary internal/external cause dis-tinction, at least not for the production of sentences inan experimental setting. Nonetheless, one might arguethat the constraints of the task led participants to pro-duce sentences that they considered to be ungrammati-cal, and that this influenced the production of reducedrelatives with internal cause verbs. This is unlikelybecause, as noted earlier, the original unreduced rela-tives could be shortened in a number of grammaticaland sensible ways. But if we assume for the moment thatit is true, then the acceptability ratings should reflect theungrammaticality, with internal cause reduced relativesjudged less acceptable than reduced relatives with allow-able external cause verbs. Again, this turns out not to bethe case. Acceptability ratings for reduced relatives withinternal cause verbs were similar to those with externalcause verbs (halfwidth = 0.4), and both were at the high


end of the 7-point scale (internal cause: M = 6.4,range = 5.3–7.0; external cause: M = 6.4, range = 4.0–7.0). There was no interaction between difficulty andverb type because verb type had roughly the same influ-ence on reduced relative acceptability ratings for the twolevels of difficulty. Reduced relative acceptability wasonly 0.2 higher for the easy external than for the easyinternal cause items (0.5), and only 0.4 higher for thehard internal than for the hard external cause items(0.5). [When the three path nominal items were dropped,mean acceptability ratings for easy internal causereduced relatives changed only from 6.5 to 6.4.] Notealso that these ratings were equivalent to those for otherconstructions produced by shortening the original unre-duced relatives, indicating that the participants did notrate the internal cause reduced relatives as less accept-able than other types of sentences (see Table 3, ‘‘OtherAcceptability’’).

Finally, the ratings for both the external cause andthe internal cause reduced relatives were similar to thosegiven to non-controversially grammatical sentences likethe original unreduced relatives (Table 3). In contrast,truly ungrammatical sentences were much less frequent,and were rated as less acceptable than the internal causereduced relatives. The percentage of shortened unre-duced relatives that were ungrammatical was 7%, andthe mean acceptability rating of these sentences was4.2. In addition, the internal cause reduced relatives werealso rated as more acceptable than grammatical butsemantically anomalous sentences like Colorless green

ideas sleep furiously, three of which were included inthe shortening study (M = 3.5) to address concernsraised by McKoon and Ratcliff (2005). Thus it also can-not be argued that participants rated meaningless inter-nal cause reduced relatives highly simply because theywere grammatically coherent.

Reduced relative acceptability ratings, especially thosefor the hard sentences, were higher in Study 2 than theywere in Study 1. The is due to the fact that in Study 1participants provided acceptability ratings for all reducedrelatives, including those intentionally designed to bedifficult to comprehend. In Study 2, on the other hand,participants had the option of creating other construc-tions, and so generally only produced reduced relativesif they considered them to be acceptable. Because partici-pants rated only their own productions in Study 2, theytended to give them high acceptability ratings.

In summary, these data clearly show that the produc-tion of reduced relatives for internal cause verbs is notprohibited—on the contrary, participants frequentlyproduced internal cause reduced relatives, and judgedthem to be as acceptable as other grammatical sentences.

Ambiguity effects

The previous section shows that a binary internal/ex-ternal cause distinction fails to make the correct predic-

tions for the shortening data. Alternatively, the difficultyin reduced relative comprehension may be due to tempo-rary ambiguity. If this is correct, then fewer reduced rel-atives should be produced for the hard sentences, whichwere intentionally designed to be difficult to interpret asa type of passive construction. This prediction wasborne out. Reduced relative probability was over threetimes greater for easy (M = 0.58) than for hard(M = 0.18) sentences (0.09). In addition, the differenceheld for both the external cause (mean difference = 0.45,halfwidth = 0.13) and the internal cause verbs (meandifference = 0.35, halfwidth = 0.13).

In addition to being more frequent, easy reduced rel-atives (M = 6.6) were also judged to be more acceptablethan hard reduced relatives (M = 6.1), showing that thefactors that increase ambiguity made the hard reducedrelatives more difficult to understand (0.4). Difficultyand verb type did not interact. Reduced relative accept-ability was 0.8 higher for external cause easy versus harditems (0.5), but there was only a 0.2 difference for theinternal cause easy versus hard items (0.5). The effectsof difficulty on the reduced relative acceptability ratingsin Study 2 are not as pronounced as in Study 1 (nor arethey as pronounced as in the proportion of producedreduced relatives) because in Study 2, participants ratedthe acceptability of only their own productions, whichby and large they considered to be highly acceptable.

Relations among constructions

A further difference between MTS and an ambiguityresolution account involves the relationship betweenunreduced and reduced relative clauses. On the MTSaccount, the two have different meanings, and as a resultinternal cause verbs are blocked from the reduced rela-tive even though they are acceptable in other passiveconstructions. On the approach taken here, reducedand unreduced relatives are highly related in meaning,but one is temporarily ambiguous whereas the other isnot. Consequently, participants should rate the two asvery similar in meaning.

The similarity ratings support the latter approach.Overall, the reduced relatives were rated as highly simi-lar in meaning to the original unreduced relatives(M = 4.7 on a scale in which 5 = the sentences have the

same meaning). In contrast, when constructions otherthan the reduced relative were generated, they were rat-ed as less similar to the original sentence than were thereduced relatives (overall M = 4.0; see Table 3 for meansby condition). The analysis of variance showed a signif-icant effect of verb type, with internal cause reduced rel-atives, M = 4.9, judged to be closer in meaning to theoriginal unreduced relatives than were the external causereduced relatives, M = 4.4 (0.4). Reduced relatives wererated as marginally more similar to the original unre-duced relatives for the easy, M = 4.8, than for the hardsentences, M = 4.4 (0.4). Furthermore, difficulty inter-


acted with verb type. Similarity ratings differed by 0.9for external cause easy versus hard items (0.6), but wereidentical for the internal cause easy and hard items.Making the comparisons in the other direction, for theeasy items, there was only a 0.1 difference between inter-nal and external cause verbs, whereas for hard items, thesimilarity ratings were 1.0 higher for internal than forexternal cause verbs.

Although the similarity ratings are consistent with anambiguity account, we believe that they should be inter-preted with caution. The means are consistently high,indicating that participants in almost all cases took thereduced relatives they created to have very much thesame meaning as the original unreduced relatives. Theprimary difference in the ratings was due to the lowermean rating for hard external cause items, and this inturn was driven by the response of a single participant,who produced The waiter served the steak enjoyed it

immensely and rated it at 1 on the 1–5 scale in similarityto the unreduced relative. This rating was then the meanfor that item. Overall, participants rated the greatmajority of the reduced relatives that they produced ashighly similar in meaning to the original unreducedrelatives.

A better method for testing relatedness across con-structions is to use the acceptability ratings for the origi-nal unreduced relatives and passives, plus the internal/external cause distinction, to predict the proportion ofreduced relatives created. In bivariate correlations, theproportion of reduced relatives correlated significantlywith unreduced relative acceptability, r = .58, and pas-sive acceptability, r = .57. However, it did not correlatewith the internal/external cause distinction, r = �.15,p > .3.

These variables were entered into a stepwise regres-sion with the probability of a reduced relative for a givensentence as the dependent variable, and acceptability ofthe unreduced relative, acceptability of the passive, andinternal/external cause status as the predictors. Unre-duced relative acceptability entered on the first step, fol-lowed by passive acceptability. Combined, these twofactors accounted for 42% of the variance in the proba-bility of producing a reduced relative, F2(2,45) = 16.02.Both variables predicted significant unique proportionsof variance: unreduced relative acceptability, r2 = .14,t2(45) = 2.64; passive acceptability, r2 = .12, t2(45) =2.53. At no time was the internal/external cause distinc-tion a significant predictor: p > .8 following step one andp > .3 following step two. The correlation and regressionanalyses together provide further evidence that a verb’sbehavior is related among passive constructions, whileproviding no evidence for the importance of a categori-cal distinction between external and internal cause verbs.Finally, note that the results were virtually identicalwhen the three path nominals with easy internal causeverbs were excluded. Together, unreduced relative and

passive acceptability accounted for 43% of the variancein the probability of producing a reduced relative,F2(2,42) = 15.88. Again, both variables predicted signif-icant unique proportions of variance: unreduced relativeacceptability, r2 = .14, t2(42) = 2.65; passive acceptabili-ty, r2 = .11, t2(42) = 2.28.

Study 3

McKoon and Ratcliff (2003) argue that Englishspeakers will not produce reduced relatives with internalcause verbs because the meaning of the reduced relativeconstruction combines with the internal cause verb tem-plate to block production (p. 506). In Study 2, by con-trast, native speakers willingly produced such reducedrelatives, and judged their own productions to be highlyacceptable. One caveat, however, is that these reducedrelatives were produced under somewhat artificial exper-imental conditions, and this may have led participants tocreate sentences that they might have otherwise avoided.

Thus it is also important to show that internal causereduced relatives are found in more naturally occurringtext, in order to provide converging evidence for theclaims made in Study 2. McKoon and Ratcliff (2003,Corpus Studies 4 and 5) searched corpora for instancesof internal cause change of state and manner of motionreduced relatives, and discovered that these do in factoccur. However, in their Corpus Study 7, they foundthat external cause reduced relatives were 100 timesmore frequent: The probability of a reduced relativewas .06 for external cause verbs, but only .0006 for inter-nal cause verbs when the two internal cause verb classeswere combined. This difference in occurrence rate wastaken as evidence for a categorical distinction betweenthe two verb classes with respect to reduced relativegrammaticality.

Interestingly, however, when the results of McKoonand Ratcliff’s (2003) Corpus Studies 4 and 5 are consid-ered individually, it becomes clear that the difference inreduced relative probability was actually larger betweenthe two classes of internal cause verbs than between theexternal cause and internal cause change of state verbs, acomparison that crosses the internal/external cause cat-egory boundary. External cause verbs were 15 timesmore likely to occur in reduced relatives than were inter-nal cause change of state verbs (probabilities of .06 and.004, respectively). This is a notable difference, butreduced relatives with internal cause change of stateverbs were in turn 27 times more frequent than withinternal cause manner of motion verbs (probability of.00015).

Although MTS offers an account of the clear distinc-tion between the external cause and internal causechange of state verbs, it fails to explain the more strikingdifference between the internal cause manner of motion


verbs, on the one hand, and the internal cause change ofstate verbs on the other. It is possible that a single factorunderlies both differences, and one likely candidate is theavailability of related constructions like the passive. Thethree verb classes do differ in this respect: All externalcause verbs may appear in the transitive (Levin, 1993)and so can occur freely in the passive. The basic senseof the internal cause change of state verbs is intransitive,but many of these also occur in the causative, and somay occur in the passive as well (McKoon & MacFar-land, 2000). In contrast, although manner of motionverbs can occur in the causative, they rarely do so, andMcKoon and Ratcliff (2003, p. 498) argue from the ‘‘rar-ity of external cause sentences for the class of manner ofmotion verbs’’ that speakers and writers follow the con-straints imposed by internal control. Although we offer adifferent account of these data in the General Discus-sion, we do agree that causatives occur infrequently withmanner of motion verbs, which are therefore also infre-quent in the passive. Thus a standard account wouldpredict that manner of motion verbs would rarely befound in the reduced relative, which is also a passiveconstruction. In the General Discussion, we describethe factors that underlie use in the intransitive, transi-tive, and passive in terms of real events in the world thatverbs (and the sentences containing them) are used todescribe.

Study 3 had two main goals. First, we replicatedMcKoon and Ratcliff’s (2003) Corpus Studies 4 and 5to verify their finding that reduced relatives are relativelyfrequent in change of state verbs, compared to verbsdescribing manner of motion. Second, we tested whetherthe probability that a verb will occur in a simple passiveis related to its use in a reduced relative. We conductedseveral regression analyses to test whether a verb’s ten-dency to occur in the reduced relative is better predictedby its probability of occurring in the transitive and thepassive, or by its status as an external or internal causeverb.

Method

Fifty-four verbs were used. These were all of theverbs used in McKoon and Ratcliff (2003, Experiment1), and McKoon and MacFarland (2002) Experiments1–3. There were 24 external cause change of state verbs,14 internal cause change of state verbs, and 16 internalcause manner of motion verbs.

Four parsed corpora were used: The Wall StreetJournal corpus and Brown Corpus (each one millionwords); Wall Street Journal 1987 (25 million words)and the British National Corpus (100 million words).The corpora differ in genre as well as size. Wall StreetJournal and Wall Street Journal 1987 are derived exclu-sively from Dow Jones newswire stories, while theBrown Corpus and the British National Corpus are

balanced corpora that include literature and news arti-cles. In addition, approximately 10% of the BritishNational Corpus consists of transcribed spoken materi-al. The Wall Street Journal corpus and Brown Corpuswere parsed as part of the Penn Treebank Project (Mar-cus, Santorini, & Marcinkiewicz, 1993). The Wall StreetJournal 1987 consists of the three year Wall Street Jour-nal collection from the Association for ComputationalLinguistics Data Collection Initiative corpus, and wasparsed using methods developed by Eugene Charniakand associates from Brown Laboratory for LinguisticInformation Processing (Charniak, 1997). All threeparsed corpora are available from the Linguistic DataConsortium at the University of Pennsylvania. The Brit-ish National Corpus is available unparsed from OxfordUniversity, and was parsed by Jeff Elman using theCharniak parser (Roland, Elman, & Ferreira, 2006.)

All sentences containing the 54 internal/externalcause verbs were extracted automatically from the fourcorpora using scripts modified from earlier studies (Hareet al., 2004; Roland, Dick, & Elman, 2006). These scriptsclassified the verbs into a set of subcategorizationframes, which were then collapsed into the more generalclassifications transitive, passive, reduced relative, andother. Sentences with an immediately post-verbal nounphrase were classified as transitive.Because some verbswith passive morphology could also be interpreted asadjectival, we followed McKoon and Ratcliff (2003,p. 507) and categorized as passive all uses of the verbthat could be interpreted that way, both for simplepassives and passive reduced relatives.

The Brown and Wall Street Journal corpora havebeen extensively hand-corrected, so their error rate isquite low. In the Wall Street Journal 1987 and BritishNational Corpus, however, the tallies of infrequent con-structions tended to be inflated relative to the other twocorpora. Therefore two raters (the first author and aresearch assistant who was unaware of the hypotheses)hand-checked all passives and reduced relatives in thosetwo corpora, and any examples that both raters consid-ered misclassified were eliminated. Any disagreementswere resolved through discussion between the two raters.

Basic search results

To compare our results with those of McKoon andRatcliff (2003, Corpus Studies 4, 5, and 7) we begin withdescriptive statistics before proceeding to analyses ofvariance and regression analyses that focus on the pro-portion of sentences that use a verb in particular struc-tures. We included only reduced relatives in which thehead is the entity participating in the activity denotedby the verb. We also eliminated all non-restrictivereduced relatives that were marked as such by punctua-tion (either commas or hyphens), as well as any thatcould solely be interpreted as an adjectival phrase


according to MTS criteria. (We are grateful to GailMcKoon for clarifying precisely which of our reducedrelatives adhered strictly to MTS criteria.)

McKoon and Ratcliff (2003, Corpus Study 4)searched for 69 internal cause manner of motion verbs,and found 6 reduced relatives out of a total of 39,159sentences. Thus they computed the probability of areduced relative sentence with a manner of motion verbas .0002 for their corpus. We searched for 16 manner ofmotion verbs, with similar results: 4 reduced relativesout of 39,000 total sentences (.0001). This occurreddespite the fact that our verb set did not include parade

or hurry, the two manner of motion verbs with 4 of the 6reduced relatives found in the earlier study. Our resultswith internal cause change of state verbs were also com-parable to those of McKoon and Ratcliff’s CorpusStudy 5. Those authors searched for 17 such verbs,and report finding 21 reduced relatives in 4775 sentences(.004). For our largely overlapping set of 14 verbs therewere 26 reduced relatives in 4429 sentences (.006).

Interestingly, although our results with internalcause verbs are similar to those reported by McKoonand Ratcliff, the proportion of reduced relatives inour external cause verbs is much lower: The proportion(reduced relatives divided by sentences) is reported as0.06 in McKoon and Ratcliff (2003, Corpus Study 7),but the comparable figure is 0.002 (87 reduced relativesin a total of 35,000 sentences) in our study. This ismost likely due to the specific verbs used. Our searchinvolved the change of state verbs used in earlier on-line internal/external cause studies by McKoon andcolleagues, which tended to be matched for overall fre-quency, whereas 73 of the 85 external cause verbs usedby McKoon and Ratcliff (2003, Corpus Study 7) werefrom previous structural ambiguity experiments thatfocused on the reduced relative. As a result, these verbswere highly transitive, with correspondingly high occur-rence in the passive. Both sets of verbs were externalcause, however, and so MTS would predict that thereduced relative construction should be equally possiblein both cases.

We suggest that the difference in reduced relative fre-quency may be related to these differences in passiviz-ability. Our external cause verbs differ from those inMcKoon and Ratcliff (2003) in their passive probabili-ties: The probability of passive occurrence for the exter-nal cause verbs in McKoon and Ratcliff (2003, p. 513)was .55, while it was .09 in the present study. If reducedrelative occurrence is related to the ability to passivize,there should be fewer reduced relatives in the presentset of verbs, and that is what was found. Further notethat the passive probabilities are similar in the two stud-ies for the internal cause change of state verbs—aswould be expected because they are largely the sameverbs—and the reduced relative probabilities are similaras well.

MTS addresses a narrowly defined set of reduced rel-ative clauses, but in general, theories of reduced relativeprocessing attempt to account for a wider range of data.For this reason, our search also included sentences inwhich, by the criteria of MTS, the head is not a partic-ipant in the activity denoted by the verb, such as in,it. . .supersedes a pitch climbed by Grant Farquhar a few

weeks previously, perhaps because the verb was used ina non-canonical sense, as in this could be done through

shares floated on the stock market. It also included sen-tences in which the reduced relative is non-restrictive,since non-restrictives may or may not be marked bypunctuation, and therefore might be difficult or notdepending on factors related to ambiguity resolution.For completeness, we note that in this set, we find amuch larger number of reduced relatives for all classesof verbs: 159 in the 35,000 external change of state sen-tences; 61 in the 4429 internal cause change of state sen-tences, and 19 in the 39,000 internal cause manner ofmotion sentences.

To this point, we have focused on comparisons of ourresults with those of McKoon and Ratcliff’s (2003) cor-pus studies. In the next section, we report inferentialtests on the probability of reduced relatives, passives,and transitives in each verb class.


This section begins with analyses of variance on theprobability of passive occurrence, P(passive) and proba-bility of transitive occurrence, P(transitive). These anal-yses show that both variables differ across external causechange of state, internal cause change of state, and man-ner of motion verbs. We then present three analysesincluding only the reduced relatives that adhere strictlyto the criteria of MTS (McKoon & Ratcliff, 2003,2005). The analysis of variance on P(reduced relative)shows that internal cause change of state reduced rela-tives have a higher probability of occurrence than eitherexternal cause change of state or manner of motionreduced relatives. We then predict P(reduced relative)using three predictors, the internal/external cause dis-tinction, P(passive), and P(transitive). P(passive) is thesole significant predictor. Next, we use a binary logisticregression analysis to predict the existence or non-exis-tence of reduced relatives for each verb. P(passive) andP(transitive) are significant predictors, whereas the inter-nal/external cause distinction again is not.

As noted above, MTS addresses a narrowly definedset of reduced relative clauses, but other theories ofreduced relative comprehension do not. Therefore it isimportant to investigate whether the same results areobtained when more general inclusion criteria areapplied—criteria that do not depend on the fine-graineddistinctions that are central to the MTS account. Forthis reason, we report the same analyses using expanded


criteria for inclusion of reduced relatives. These analysesalso include sentences in which the head is not a proto-typical participant in the activity denoted by the verb(such as path nominals with manner of motion verbs),and those in which the reduced relative is potentiallynon-restrictive. We did, however, continue to eliminatecases that could solely be interpreted as an adjectivalphrase according to MTS criteria. The analyses of vari-ance show a significant main effect of verb class, withinternal cause change of state verbs again having thehighest P(reduced relative). When predicting P(reducerelative), P(passive) and the internal/external cause dis-tinction are significant predictors, although internalcause reduced relatives are predicted to have a higherprobability of occurrence, contrary to MTS. The binarylogistic regression again shows that P(passive) andP(transitive), but not the internal/external cause distinc-tion, predict the existence of reduced relative clauses.

Probability of passives and transitives

The proportion of each type of construction was cal-culated as the number of times each verb appeared in aspecific structure divided by the number of times thatany morphological form of the verb appeared in the cor-pus. Thus, for the first two analyses of variance, thedependent variables were the proportion of passivesand proportion of transitives. The independent variablewas verb subtype, which had three levels: external causechange of state, internal cause change of state, and inter-nal cause manner of motion. Verb subtype was betweenitems (F2). Mean passive, transitive, and reduced relativeprobabilities for each verb subtype are presented inTable 5, and the analyses of variance statistics are pre-sented in Table 6.

For P(passive), there was a main effect of subtype.Passive probability was .0747 higher for internal causechange of state than for manner of motion verbs(.0559). It was .0839 higher for external cause changeof state than for manner of motion verbs (.0493). The0.0092 difference was not reliable for internal cause ver-sus external cause change of state verbs (0.0538).

Table 5Proportions of each type of construction for Study 3 corpora analyse

Construction External cause change ofstate

I

M SE

Passive 0.090 0.014 0Transitive 0.205 0.018 0

MTS Criteria

Reduced relative 0.0051 0.0020 0

Expanded Criteria

Reduced relative 0.0075 0.0024 0

For P(transitive), there was also a main effect of sub-type. Transitive probability was .0898 higher for exter-nal cause change of state verbs than for the manner ofmotion verbs (.0662). Internal cause change of stateverbs had a marginally higher P(transitive) than mannerof motion verbs, with a difference of 0.0724 (0.0751).The 0.0174 difference was not reliable for internal causeversus external cause change of state verbs (0.0690).

We note that Merlo and Stevenson (1998), who con-ducted a similar analysis, found similar results. In partic-ular, in their data, the probability of an intransitive usewas significantly higher for manner of motion verbs(their unergatives) than for external cause change ofstate (their unaccusatives).

Analyses using MTS criteria

The following set of analyses includes only reducedrelatives that meet MTS inclusion criteria (McKoon &Ratcliff, 2003, 2005).

Probability of reduced relatives

There was a main effect of subtype. Reduced relativeprobability was .0143 higher for internal cause change ofstate than for manner of motion verbs (.0114). Theprobability of a reduced relative was marginally higher,with a difference of .0093, for internal cause change ofstate than for external cause change of state verbs(.0105). The .0050 advantage for external cause changeof state over manner of motion verbs was not reliable(.0101). Therefore, using MTS criteria for inclusion ofreduced relatives, the corpus data show that reduced rel-atives with internal cause verbs are clearly not prohibit-ed. This result is consistent with the production data inStudy 2.

In summary, no categorical difference was found inreduced relative occurrence rate between internal andexternal cause change of state verbs, contrary both tothe predictions of MTS and to the corpus studies pre-sented in McKoon and Ratcliff (2003). The currentresults also show that the change of state verbs in ourcorpus search are evenly matched on their tendency to

s

nternal cause change ofstate

Internal cause manner ofmotion

M SE M SE

.081 0.023 0.006 0.003

.188 0.024 0.115 0.019

.0144 0.0059 0.0001 0.0001

.0241 0.0083 0.0003 0.0001

Table 6Analyses of variance for Study 3 corpora analyses

Dependent variable Effect Items

Probability passive Verb type F(2,51) = 9.22Probability transitive Verb type F(2,51) = 5.61

MTS Criteria

Probability reduced relative Verb type F(2,51) = 4.68

Expanded Criteria

Probability reduced relative Verb type F(2,51) = 7.26

Note. Verb type = external cause change of state vs. internal cause change of state vs. internal cause manner of motion; all F’s aresignificant.


occur in the passive. This was not the case for the changeof state verbs in the McKoon and Ratcliff (2003) corpusanalyses, and might explain why the reduced relative—another passive construction—occurred more frequentlywith external cause than with internal cause verbs in thatstudy. This is also consistent with our finding that man-ner of motion verbs, which have a lower probability of apassive than the change of state verbs, differ from thoseverbs in their reduced relative probability as well.

Our analyses of the verbs from McKoon and col-leagues’ on-line studies suggests that reduced relativesare not prohibited with internal cause verbs. It also sug-gests that the relevant distinction among these three verbtypes is change of state versus manner of motion,because of the differential availability of the passive inthe three verb types. In the next set of analyses, we testmore directly whether the probability of the passive pre-dicts a verb’s tendency to occur in the reduced relativeconstruction.

Factors predicting reduced relative probability

We first tested whether the probability of a reducedrelative for each verb, P(reduced relative), correlateswith the probability of a passive P(passive), probabilityof a transitive P(transitive), and the categorical variableinternal/external cause class (coded as externalcause = 0, internal cause = 1). When each variable wasregressed separately against P(reduced relative), P(pas-sive) predicted 21% of the variance, F(1,52) = 13.64,and P(transitive) predicted 13% of the variance,F(1,52) = 7.67. In contrast, internal/external cause verbclass predicted less than 1%, F < 1.

We then entered these variables into a stepwiseregression with P(reduced relative) as the dependent var-iable, and P(passive), P(transitive), and internal/externalcause as the predictors. P(passive) was the sole signifi-cant predictor, and thus the percent variance accountedfor and its significance are the same as in the zero-orderanalyses reported in the previous paragraph. AfterP(passive) entered the equation, P(transitive) was non-significant, p > .2. The internal/external distinction wasa marginal predictor, with a partial correlation of .26,

t(52) = 1.92, p = .061. However, contrary to the predic-tions of MTS, this correlation is positive, indicating thatthere is a higher probability of reduced relatives in theinternal cause than in the external cause verbs whenP(passive) is already in the equation. In addition, asshown above, the internal/external cause distinctiondoes not correlate with P(reduced relative) on its own.

Factors predicting the existence of reduced relatives

In the previous section, the dependent variable wasthe probability of a reduced relative. MTS, however,makes categorical predictions regarding the existenceor non-existence of reduced relatives for certain classesof verbs. Therefore, it could be argued that a moreappropriate test of the MTS account is a logistic regres-sion in which the dependent variable represents whetheror not reduced relatives were found for each verb (i.e.,reduced relatives do not exist = 0, reduced relativesexist = 1). Overall, there was at least 1 reduced relativethat fit the MTS criteria for 21 of the 54 verbs. Twelveof the 24 external cause change of state verbs, and 9 ofthe 30 internal cause verbs, including 7 of the 14 internalcause change state verbs, and 2 of the 16 manner ofmotion verbs, occurred in the reduced relative.

We conducted a binary logistic regression usingP(passive), P(transitive) and verb type (internal/externalcause) as predictors. The forward conditional methodwas used in SPSS for conducting stepwise logistic regres-sions. The model began with a baseline prediction rateof 61%. Because 33 of the 54 verbs did not occur inreduced relatives, the baseline prediction was that noverbs occur in reduced relatives (thus 33/54 · 100 = 61% correct predictions).

P(transitive) and P(passive) were the only significantpredictors of the existence of reduced relatives. P(transi-tive) entered first, Nagelkerke R2 = .30, v2(1) = 13.53,and correctly assigned 42 of the 54 verbs (78%). P(pas-sive) entered next, increasing the prediction rate to 43of the 54 verbs (80%), with Nagelkerke R2 = .39,v2(1) = 4.92. The LR statistic, which indicates how themodel’s predictions would change if a specific variablewas removed, showed that, at this point, P(transitive)


was a significant predictor, LR(1) = 5.01, as was P(pas-sive), LR(1) = 4.44. At no step was internal/externalcause a significant predictor of the existence/non-exis-tence of reduced relatives, all p’s > .7.

In summary, this pair of regression analyses showthat a verb’s probability of appearing in the passive isthe dominant predictor of its probability of appearingin a reduced relative when MTS criteria for the inclusionof reduced relatives is used. Furthermore, the verb’sprobability of appearing in the transitive also was a sig-nificant predictor of the existence of reduced relatives. Inno analysis did the internal/external cause distinctionaccount for the reduced relative production in corpora.

Analyses using expanded criteria

The following analyses use somewhat expanded crite-ria for including reduced relatives, ones that do not fol-low all of the MTS criteria, as outlined above.

Probability of reduced relatives

There was a main effect of subtype. Reduced relativeprobability was .0238 higher for internal cause change ofstate than for manner of motion verbs (.0154). Theprobability of a reduced relative was .0166 higher forinternal cause change of state than for external causechange of state verbs (.0142). The .0072 advantage forexternal cause change of state over manner of motionverbs was not reliable (.0136). Therefore, using theexpanded criteria for inclusion of reduced relatives, thecorpus data again show that reduced relatives with inter-nal cause verbs are clearly not prohibited. This result isconsistent with the production data in Study 2, as well asthe analysis of variance using the MTS criteria.

Factors predicting reduced relative probability

We again tested whether P(reduced relative) corre-lates with P(passive), P(transitive), and the categoricalvariable internal/external cause class. We first used eachvariable on their own. P(passive) predicted 26% of thevariance in P(reduced relative), F(1,52) = 18.62. P(tran-sitive) predicted 12% of the variance in P(reduced rela-tive), F(1,52) = 6.80. Internal/external cause verb classpredicted a non-significant 1% of the variance, F < 1.

We then entered these variables into a stepwiseregression with P(reduced relative) as the dependent var-iable, and P(passive), P(transitive), and internal/externalcause as the predictors. P(passive) entered first, as theresults in the previous paragraph indicate. Internal/ex-ternal cause entered on the second step. Its partial corre-lation was r = .34, t(51) = 2.58, with P(passive) having apartial correlation of r = .59, t(51) = 5.15. Thus, inter-nal/external cause is a significant predictor when the var-iance due to P(passive) is removed. Contrary to thepredictions of MTS, however, this correlation is posi-tive, indicating that there is a higher probability of

reduced relatives in the internal cause than in the exter-nal cause verbs when P(passive) is already in the equa-tion. In addition, as shown above, the internal/externalcause distinction does not correlate with P(reduced rela-tive) on its own. P(transitive) was not significant at anystep.

Factors predicting the existence of reduced relatives

Overall, there was at least 1 reduced relative for 33of the 54 verbs. Eighteen of the 24 external cause verbsand 15 of the 30 internal cause verbs, including 11 ofthe 14 internal cause change state verbs, and 4 of the16 manner of motion verbs, occurred in the reducedrelative.

We again conducted a logistic regression using P(pas-sive), P(transitive) and verb type (internal/externalcause) as predictors. The model began with a baselineprediction rate of 61%. Because 33 of the 54 verbsappeared in reduced relatives, the baseline predictionwas that all verbs occur in reduced relatives (thus 33/54 · 100 = 61% correct predictions).

P(transitive) and P(passive) were the only significantpredictors of the existence of reduced relatives. P(transi-tive) entered first, Nagelkerke R2 = .41, v2(1) = 19.41,and correctly assigned 39 of the 54 verbs (72%). P(pas-sive) entered next, increasing the prediction rate to 43of the 54 verbs (80%), with Nagelkerke R2 = .52,v2(1) = 6.81. The LR statistic showed that, at this point,P(transitive) was a significant predictor, LR(1) = 7.37,as was P(passive), LR(1) = 7.51. At no step was inter-nal/external cause a significant predictor of the exis-tence/non-existence of reduced relatives, all p’s > .4.

The regression analyses show that a verb’s probabil-ity of occurrence in the passive is the most consistentpredictor of whether it will occur in the reduced relativeconstruction. This is expected on any account in whichthe reduced relative is a type of passive construction,and difficulty with the reduced relative arises when fac-tors conspire to make the head noun difficult to interpretas the patient of the passive verb. On such accounts,reduced relatives should be more common for verbs witha higher passive probability, and the results of theregression models bear this out. In addition, P(transi-tive) was a significant predictor of the existence ofreduced relatives both when the MTS and the expandedinclusion criteria were used. Transitive usage is onedeterminant of whether a passive construction can beformed, and thus whether a reduced relative might beused. Finally, the internal/external cause distinctionfailed to be a significant predictor of the existence ornon-existence of reduced relatives. When the probability(rather than the existence) of reduced relatives was thedependent variable in the expanded criteria analyses,the internal/external cause distinction was a significantpredictor, but in the direction opposite to the predictionsof MTS.


Study 3 found that reduced relatives are more com-mon for change of state than for manner of motionverbs, and that the probability that a change of stateverb occurs in a reduced relative is similar for both inter-nal and external cause verbs. This contradicts the claimsof MTS, but is consistent with the results of previouscorpus studies. As mentioned earlier, McKoon and Rat-cliff (2003) found a similar difference between internalcause change of state and manner of motion verbs intheir own corpus. Merlo and Stevenson (1998) foundthe same pattern when they contrasted three sets ofverbs: Unergatives (internal cause manner of motion),unaccusatives (external cause change of state) and ‘‘ob-ject-drop’’ verbs. As in our corpus, the unaccusative(change of state) verbs were much more frequent inthe passive and transitive than were the unergative(manner of motion) verbs, and showed a corresponding-ly higher probability of occurrence in the reduced rela-tive. The authors noted that significant differences onthese and related dimensions paralleled speakers’ intui-tive judgments of comprehension difficulty.

The third set, so-called ‘‘object-drop’’ verbs, wereexternal cause verbs from a variety of semantic sub-clas-ses, as were those in McKoon and Ratcliff’s (2003) Study7. These (like McKoon & Ratcliff’s external cause verbs)had a much higher proportion of transitive and passiveuses than did their external cause change of state verbs,and a higher probability of reduced relatives as well.This is consistent with our results, and, in addition, sup-ports the argument that McKoon and Ratcliff foundmore external cause reduced relatives than we didbecause the verbs they searched for had a higher proba-bility of passive occurrence.

General discussion

The results of the current studies fail to support any ofthe empirical claims made by the Meaning through Syn-tax theory of the reduced relative construction, as artic-ulated in McKoon and Ratcliff (2003, 2005). Englishspeakers readily produce reduced relatives with internalcause verbs, and their ease of comprehension is basedon factors that facilitate ambiguity resolution, such asthe goodness of the head as patient of the verb and thepresence of a postverbal prepositional phrase. This istrue whether or not the head noun phrase is a prototyp-ical participant in the activity. Furthermore, the best pre-dictor of whether or not a verb will occur in a reducedrelative is its probability of occurring in the transitiveand in the passive more generally. These findings areinconsistent with the MTS claim that the difficulty incomprehending sentences such as The horse raced past

the barn fell results from the ungrammaticality of internalcause verbs in the reduced relative construction. On thecontrary, our results show that many reduced relatives

with internal cause verbs are acceptable and easy to com-prehend, whereas many with external cause verbs are dif-ficult to comprehend, and rated as less acceptable. Onemight argue that nothing in MTS excludes difficulty withexternal cause reduced relatives: These are predicted tobe grammatical, but a variety of factors contributing toholistic meaning may then influence how easy they areto comprehend (McKoon & Ratcliff, 2005). However,appealing to holistic meaning, which was defined in onlyvery general terms by McKoon and Ratcliff (2005), doesnot explain why reduced relatives with internal causeverbs were rated to be as comprehensible as those withexternal cause verbs (Study 1). Perhaps more important-ly, it does not account for the fact that English speakersproduced numerous reduced relatives with internal causeverbs (Study 2). Even if the graded acceptability ofreduced relatives with external cause verbs could beexplained by a notion of holistic meaning, it is unclearwhat additional contribution is offered by MTS itself,and how holistic meaning might offer a more satisfactoryaccount than existing, more clearly articulated con-straint-based models (MacDonald et al., 1994; McRaeet al., 1998; Tanenhaus & Trueswell, 1995).

Stated differently, there are two fundamental weak-nesses with the MTS account of the production andcomprehension of the reduced relative construction.First, the variables that best explain the data are notincorporated into the MTS account, whereas they arecentral to other current views of the language processingsystem, whether two-stage or constraint-based. More-over, incorporating those variables would seem to beinconsistent with the basic tenets of the MTS program(McKoon & Ratcliff, 2003). Second, the factor that iscentral to MTS, the dichotomous classification of verbsinto those that do or do not have an external cause intheir event template, does not account for a significantportion of the variance.

Could the MTS account of reduced relatives berevised to accommodate the current results? McKoonand Ratcliff (2003, p. 498) do note that internal causalitycan be overridden. MTS could incorporate this into thetheory by positing that internal cause verbs also have anexternal cause template, which would be the basis forcausative uses such as the boy walked his dog or winter

storms eroded the beach. The advantage of this approachis that it would allow these verbs to avoid the prohibi-tion on internal causality, and also capture the fact thatinternal cause reduced relatives are possible in preciselythose cases where the causative is also allowed.

McKoon and Ratcliff (2003) consider and reject thispossibility both for change of state (p. 502) and mannerof motion verbs (p. 498). And indeed, it would seriouslyundermine the empirical support for MTS, in particularthe interpretation of the reading time and lexical deci-sion results in McKoon and Ratcliff (2003, 2005) andMcKoon and MacFarland (2000, 2002). In these studies,


differences in reaction times to internal and externalcause verbs have been taken to indicate differences inprocessing time due to the greater complexity of theexternal cause template. This interpretation would notbe possible if an external cause template were availableto the internal cause verbs as well, because that wouldeliminate the difference in template complexity. Henceit does not appear that MTS could permit any straight-forward modification of the single-template assumption,even if it would allow MTS to accurately account for theease or difficulty of reduced relatives.

Alternatively, the MTS program could abandon theeffort to explain the difficulty of some reduced relatives.This could be done as part of a more general decision torestrict the MTS program to explaining verb-based com-plexity differences in the comprehension of unambigu-ous sentences, and the production patterns associatedwith different classes of verbs in different syntactic con-structions. Given the well-documented systematic rela-tionship among the type of event or activity that averb denotes, the nature of the entities that participatein that event, and the types of syntactic complementswith which that verb can occur, exploring the interactionof verb-type and construction is likely to lead to impor-tant insights. However, as we discuss below, we antici-pate that the assumptions that compromise the MTSaccount of the reduced relative construction, in particu-lar the assumption that each verb has only one eventtemplate, are likely to prove problematic for MTSaccounts of the comprehension and production of otherverb types in other constructions.

In the remainder of the article, we focus on thebroader implications of our results. We begin by consid-ering the dangers of drawing certain types of inferencesabout acceptability from frequency of occurrence in cor-pora. We then focus on the relationships among the typeof event or activity that a verb denotes, the nature of theentities that participate in that event, and the types ofsyntactic complements with which that verb can occur.

Estimating occurrence from corpora

McKoon and Ratcliff (2003) point out advantages ofcombining experimental studies with corpus analyses,and we concur. Although both methods are useful, eachhas its strengths and weaknesses. In experimental stud-ies, the experimenter can carefully construct materialsto test specific hypotheses. However, these materialsare unlikely to be representative of typical languageand the use of somewhat artificial stimuli and tasksmay bias the results. Corpus analyses, on the otherhand, involve more natural and realistic language useand can provide valuable insights about frequency ofoccurrence, as well as patterns of co-occurrence. Howev-er, the relevant data may be sparse, with results that arelikely to vary depending on the corpus.

It is important to keep in mind that even the largestcorpus represents a relatively small sample of the record-ed exemplars of the sentences and utterances of a lan-guage. Any corpus or composite of corpora is limited,and may not contain data that exists elsewhere. Thusone must keep in mind the probabilistic nature of locat-ing rare constructions like the reduced relative, giveninfrequent verbs and a restricted language sample (cf.Tomasello & Stahl, 2004).

The sparseness problem becomes particularly acutewhen drawing conclusions from the absence of particu-lar classes of exemplars. McKoon and Ratcliff’s (2003,2005) crucial evidence for a prohibition on reduced rel-atives for verbs of internal causality was that suchreduced relatives were rare in their corpora. However,there are a least two reasons to be wary of drawing suchconclusions from negative corpus results (in essence, anull effect). First, the frequency of a construction in acorpus needs to be evaluated relative to its expectedoccurrence. McKoon and Ratcliff (2003, Studies 4 and5) do find reduced relatives with internal cause verbs,but dismiss them as errors because they are so rare.However, when one considers how infrequently theseverbs occur in passive constructions in which (accordingto MTS) they are not prohibited, then the observedprobability of their occurring in a reduced relative is,in fact, consistent with the expected probability.

Second, although both our corpus and that of McK-oon and Ratcliff (2003) are quite large, they still representonly a limited sample of the language. Hence the fact thata particular low-frequency verb is not found in a reducedrelative in one corpus or the other is, at best, extremelyweak evidence that such reduced relatives do not exist.Drawing this conclusion is akin to accepting the nullhypothesis, based on negative results with a small sampleand no power estimate. And indeed, McKoon and Rat-cliff’s corpus contained reduced relatives with the inter-nal cause verb deteriorate, although ours did not;conversely, our corpus, but not theirs, contained internalcause reduced relatives with blister and wither. Thus thelack of evidence for internal cause reduced relatives, ina particular corpus, cannot be taken as evidence that theyare either prohibited or unacceptable. Corpus data canserve as a source of hypotheses, but these hypothesesneed to be evaluated by behavioral data, including judg-ments. Our general point here, of course, simply echoesChomsky’s classic arguments about the limitations ofbuilding a linguistic theory whose goal is to account forobservations from corpora (Chomsky, 1957, 1965).

Events, verb semantics, and structure

Sense and structure

An appealing aspect of the Meaning Through Syntaxapproach is that it incorporates formal apparatus fromlexical semantics as a way of implementing the insight


that a verb’s semantic representation is reflected in, andat least partially determines, the syntactic frames inwhich the verb appears. MTS predicts that the relation-ship is straightforward, such that ‘‘the syntactic posi-tions in a sentence are defined, at least in major part,in terms of the semantic event templates’’ (p. 495). Thus,for example, in the sentence John broke the window,

‘‘John being the subject of the sentence conveys thatsomething John did is the immediate cause of the break-ing event and the window being the direct object conveysthat the window changes state as a result of the event’’(McKoon & Ratcliff, 2003, p. 495). Furthermore, MTStakes the novel step of combining verb-based event-tem-plates with Construction Grammar, which argues thatgrammatical constructions like the passive or the transi-tive make an independent contribution to meaning.

We want to emphasize that the problems with theMTS model of the reduced relative are not due to anyproblems with the ideas from construction grammarand lexical verb semantics that MTS draws upon. Rath-er they come from claims that are specific to MTS. Themost problematic assumption—one that is crucial to theMTS account of reduced relatives—is that with theexception of truly ambiguous verbs, each verb has onlya single event template, which remains invariant acrossdifferent senses and syntactic frames. This assumptioncontributes to MTS’s inability to account for reducedrelative use, and the problems become even more evidentwhen MTS is faced with a broader range of data thanthe reduced relative construction. Most verbs have dif-ferent senses, and these tend to occur in different struc-tural frames, and theories of lexical semantics take itas a major goal to account for such alternations (cf.Pustejovsky, 1995). Consider the verb race. Its internalcause template correctly predicts that race will occur inintransitive sentences like the horse raced. It also correct-ly predicts that the horse will occur in subject position,because that is the grammatical role filled by the internalcause of the event. However, this template does not pre-dict that the verb will occur in transitive sentences likeJohn raced the horse. Nor can it account for the fact John

surfaces as subject. Instead, this entity is said to comefrom other parts of the verb’s meaning (McKoon & Rat-cliff, 2003, p. 496), not from the syntactically relevantevent template (p. 494). Thus the one-template stipula-tion not only leads to the wrong structural predictions,but it also undermines the essential notion of MTS, that‘‘the event template meaning of the verb is expressedthrough the syntax of the sentence’’ (p. 495).

Clearly, if all things were equal, it would be desirablefor all instances of a verb to have a single event represen-tation. However, theories of lexical semantics must bal-ance the desire to unite multiple verb senses with theneed to account for structural alternations. The require-ment to handle a broader range of data leads theories oflexical semantics to take a more nuanced approach than

the one verb, one event template approach adopted byMTS. As one example, Levin and Rappaport Hovav(1995) propose lexical rules that alter the causal statusof the intransitive subject in specific contexts. Morecompellingly, Pustejovsky (1995) argues that the verb’srepresentation itself does not change, but the senses itwill express (and the structures it will occur in) derivefrom a generative process combining the verb’s lexicalrepresentation with those of specific arguments withwhich it co-occurs in each sense. Although this approachhas not yet been fully tested, it indicates the richness ofthe representational system required to capture the com-plex relationship between structure and meaning.

Event templates and event representations

The MTS account assumes that use of a verb in a giv-en construction is determined by its event template. Onan ambiguity account, the interpretability of a reducedrelative is related to the ease with which the initial nounphrase can be construed as the patient of the passive—ormore generally, as affected by the event. For a wide vari-ety of cases, this is equivalent to saying that the eventmust be interpretable as externally caused. However,the interpretation is influenced by factors that do notalways align with verb class. In what follows, we sum-marize these factors and their relationship to the inter-nal/external cause distinction.

First, a verb is likely to occur in the passive only if itis acceptable in the transitive. As a result, different verbclasses will vary in their passivizability. As noted earlier,external cause verbs are transitive in their basic sense,and so will passivize freely. Internal cause verbs, if theyare used in a causative sense, will occur in the transitiveas well. Pinker (1989) investigates the conditions oncausativization, and finds that individual verbs differ intheir degree of inherent internal cause, even though theymay belong to the same semantic class. As a result, thelikelihood that a given type of event will be construedas externally caused falls on a continuum, and as a gen-eral a rule, the internal and external cause verbs will befound at its extremes. However, this is a probabilisticdistinction, not a categorical one. Thus although man-ner of motion verbs like totter or scurry are rare in thecausative, they will be used that way if the movementis construed as externally induced, as McKoon and Rat-cliff (2003) find in their corpora with examples like she

tottered him home after one of his binges or . . .she had

some rather unusual things to say about CNN, and we

. . .very quickly scurried her off.Second, the causative requires an agent that is capa-

ble of inducing the action. McKoon and Ratcliff (2005)find that in the majority of cases, the grammatical sub-jects of manner of motion verbs are humans, actionsby humans, or natural forces (p. 1034), but the optionsare much wider when the verb is an external causechange of state. This is offered as evidence for a categor-


ical distinction between internal and external causeverbs, but in fact it follows from more general aspectsof thematic fit, and as such it again reflects probabilitiesrather than a rigid distinction. Human agents (or theiractions) are more likely than artifacts or abstract entitiesto cause an animate entity to move, as manner of motionverbs generally require. But manner of motion verbs arealso used in non-literal senses that do not require phys-ical force, and thematic fit extends to cases of this sort,which McKoon and Ratcliff acknowledge as problemat-ic for MTS: ‘‘For example, in An epilogue jumps us from

April to August, with the manner-of-motion verb jump

(from McKoon & Ratcliff’s, 2003, corpus), the restric-tions listed in Table 1 are violated because an epilogueis not a person, an act by a person, or a natural force.’’(2005, p. 1035).

Third, seminal work by Lakoff (1977) and Hopperand Thompson (1980) argues that transitivity is basednot simply on the verb, but on characteristics of theclause as a whole, such as the number of participants,the volitionality of the agent, and the degree to whichthe patient is construed as affected. Passivizability isdirectly related to these factors (Rice, 1987), particularlyto the interpretation of the patient’s role. Crucially, thisaccount acknowledges that locations or paths are gener-ally not treated as direct objects, yet if they are con-strued as affected by an event—if the event is habitualin that location, or otherwise characterizes it as spe-cial—they are interpreted as participants in the eventand one finds acceptable passives, as in That peak was

first climbed by Hillary in 1952 (Bolinger, 1975, 1977;Rice, 1987).

These constraints translate directly to the findings onreduced relative probability. Highly transitive verbs,particularly those with human agents, are more likelyto occur in the passive, and therefore in the reduced rel-ative (McKoon & Ratcliff, 2003; Merlo & Stevenson,1998). Change of state verbs, whether external cause likebreak or internal cause like erode, will occur in thereduced relative if the change is viewed as caused orinduced by some other entity. Manner of motion verbs,which are the least likely to be construed as induced, arenot surprisingly the least likely to be found in thereduced relative: The majority of manner of motionverbs encode highly specific movements that are difficultto construe as caused by some entity other than themover. Those that do occur in the causative, however,tend to have animate, volitional agents (McKoon &Ratcliff, 2005) or highly affected patients, two factorsthat the work cited above has shown to facilitate occur-rence in passive constructions like the reduced relative.Finally, even non-prototypical patients like locationswill appear as subject of the passive if they are construedas participants in the event. These can then appear as thehead of a corresponding reduced relative, as in ourexample the storm. . .hurled rocks and boulders onto the

path walked by tourists, where the habitual use of thatpath by tourists marks it as distinct.

Thus, whereas semantic factors largely determinewhether or not a verb may occur in the reducedrelative, these factors go beyond the information thatcould be encoded in the event template of individualverbs. It is possible that these factors could be capturedin a more richly articulated and dynamic model oflexical representation, perhaps along the lines sketchedby Pustejovsky (1995) and Jackendoff, among others.However, we would argue that a complete account ofreduced relative comprehension and use will includeconsiderations that extend beyond the lexical semanticsof specific verbs, to the multitude of factors thatinfluence how the language user construes the event beingdescribed. On this view, the relevant generalizations arenot strictly about lexical knowledge, but rather the speak-ers’ interpretation of generalized events in the world.

Appendix A

Study 1 and 2 Sentences: Reduced relatives, unreducedrelatives, and full passives, along with their mean acceptabilityrating

Easy sentences with external cause verbs

Most of the applicants interviewed by the catering companywere offered jobs. 6.82Most of the applicants that were interviewed by the cateringcompany were offered jobs. 6.10Most of the applicants were interviewed by the cateringcompany. 6.42

The evidence reviewed by the judge turned out to be unreliable.6.36The evidence that was reviewed by the judge turned out to beunreliable. 5.90The evidence was reviewed by the judge. 7.00

The suspect detained for questioning was later released. 6.64The suspect who was detained for questioning was laterreleased. 6.40The suspect was detained for questioning. 7.00

Most of the intelligence evaluated by the subcommittee wasbadly flawed. 4.83Most of the intelligence that was evaluated by the subcommitteewas badly flawed. 5.45Most of the intelligence was evaluated by the subcommittee.4.70

Whiskey aged for less than a decade often has a harsh taste.6.58Whiskey that has been aged for less than a decade often has aharsh taste. 6.27The whiskey was aged for less than a decade. 6.00

The raccoons examined by the vet were confused but not rabid.5.36The raccoons that were examined by the vet were confused butnot rabid. 4.90The raccoons were examined by the vet. 6.75


The students caught cheating on exams were suspended for atleast a semester. 6.25The students who were caught cheating on exams weresuspended for at least a semester. 6.18The students were caught cheating on exams. 7.00

The POW tortured by his captors eventually escaped.5.50The POW who was tortured by his captors eventually escaped.6.45The POW was tortured by his captors. 6.70

The movie directed by Robert Redford won three Academyawards. 6.50The movie that was directed by Robert Redford won threeAcademy awards. 6.75The movie was directed by Robert Redford. 7.00

The customer served by the efficient and good-natured waiterleft him a big tip. 5.30The customer who was served by the efficient and good-naturedwaiter left him a big tip. 5.42The customer was served by the efficient and good-naturedwaiter. 6.18

The tasks assigned to the secretary kept her busy all day.6.50The tasks that were assigned to the secretary kept her busy allday. 6.83The tasks were assigned to the secretary. 6.91

The murderer sentenced to life in prison was never given parole.5.80The murderer who was sentenced to life in prison was nevergiven parole. 6.42The murderer was sentenced to life in prison. 6.55

Hard sentences with external cause verbs

The bicycle smashed into the wall had been stolen. 3.36The bicycle that had been smashed into the wall had beenstolen. 5.10The bicycle was smashed into the wall. 6.25

The tree snapped in half during the storm took down a powerline. 3.09The tree that had been snapped in half during the storm tookdown a power line. 6.10The tree was snapped in half during the storm. 7.00

The lioness hunted throughout the night was pregnant withcubs. 3.55The lioness that was hunted throughout the night was pregnantwith cubs. 5.40The lioness was hunted throughout the night. 6.17

The mailman handed a letter put it in his bag. 2.55The mailman who was handed a letter put it in his bag.5.80The mailman was handed a letter. 6.92

The general expected to win the battle was soundly defeated.3.75The general who was expected to win the battle was soundlydefeated. 5.73The general was expected to win the battle. 6.70

The motorcycle crashed into a fire hydrant will be expensive torepair. 3.42The motorcycle that had been crashed into a fire hydrant will beexpensive to repair. 5.18The motorcycle was crashed into a fire hydrant. 6.30

The patient refused treatment for cancer sued for damages. 3.50The patient who had been refused treatment for cancer sued fordamages. 6.82The patient was refused treatment for cancer. 5.80

The judge denied parole served time. 3.42The judge who was denied parole served time. 4.64The judge was denied parole. 5.30

The waiter served a steak enjoyed it immensely. 1.70The waiter who was served a steak enjoyed it immensely. 5.92The waiter was served a steak. 4.18

The employer sent flowers as a gesture of thanks just lovedthem. 1.80The employer who was sent flowers as a gesture of thanks justloved them. 5.08The employer was sent flowers as a gesture of thanks. 6.82

The guard searched inside the prison walls had cocaine hiddenin his jacket. 2.20The guard who was searched inside the prison walls had cocainehidden in his jacket. 5.25The guard was searched inside the prison walls. 5.18

The general presented an ultimatum surrendered. 1.90The general who was presented an ultimatum surrendered. 5.75The general was presented an ultimatum. 6.55

Easy sentences with internal cause verbs

The gunshot victims rushed to the hospital were placed on lifesupport. 5.27The gunshot victims who were rushed to the hospital wereplaced on life support. 6.20The gunshot victims were rushed to the hospital. 7.00

The beaches eroded by the storm were closed to the public. 6.18The beaches that had been eroded by the storm were closed tothe public. 6.30The beaches were eroded by the storm. 7.00

Whiskey fermented in oak barrels can have a woody taste. 5.73Whiskey that has been fermented in oak barrels can have awoody taste. 5.70Whiskey is fermented in oak barrels. 7.00

Dogs walked frequently are usually well behaved. 5.64Dogs that are walked frequently are usually well behaved. 6.50Dogs are walked frequently. 6.75

The flowers wilted by the midday sun recovered in the cool ofthe night. 5.33The flowers that had been wilted by the midday sun recoveredin the cool of the night. 5.91The flowers were wilted by the midday sun. 6.70

The path traveled by many settlers extended far to the west. 6.00The path that was traveled by many settlers extended far to thewest. 6.09The path was traveled by many settlers. 6.80


The seeds germinated on the windowsill were killed by an earlyfrost. 4.83The seeds that had been germinated on the windowsill werekilled by an early frost. 5.45The seeds were germinated on the windowsill. 5.60

The silver tray tarnished by damp air was restored by theantique dealer. 5.58The silver tray that had been tarnished by damp air wasrestored by the antique dealer. 6.64The silver tray was tarnished by damp air. 5.60

The sinks corroded by the dripping water were eventuallyreplaced. 6.00The sinks that had been corroded by the dripping water wereeventually replaced. 6.33The sinks were corroded by the dripping water. 6.64

The mountain climbed by the tourists sloped gently upward.5.40The mountain that was climbed by the tourists sloped gentlyupward. 5.83The mountain was climbed by the tourists. 6.18

The city streets roamed by gangs of young men were toodangerous for tourists. 5.60The city streets that were roamed by gangs of young men weretoo dangerous for tourists. 5.42The city streets were roamed by gangs of young men. 6.27

Canadian cars rusted by road salt don’t last as long as theyshould. 5.10Canadian cars that have been rusted by road salt don’t last aslong as they should. 6.50Canadian cars are rusted by road salt. 5.82

Hard sentences with internal cause verbs

The worm wiggled in front of the fish was irresistible bait.3.82The worm that was wiggled in front of the fish was irresistiblebait. 5.30The worm was wiggled in front of the fish. 5.42

The horse raced past the barn fell. 2.64The horse that was raced past the barn fell. 2.80The horse was raced past the barn. 5.33

The raft floated down the river sank. 3.27The raft that was floated down the river sank. 4.60The raft was floated down the river. 5.67

The princess waltzed across the dance floor waved at the band.3.18The princess who was waltzed across the dance floor waved atthe band. 5.10The princess was waltzed across the dance floor. 6.08

The woman walked through the park every day was sufferingfrom Alzheimer’s disease. 3.83The woman who was walked through the park every day wassuffering from Alzheimer’s disease. 4.55The woman was walked through the park every day. 4.20

The pony galloped past the crowd held its head high. 3.92The pony that was galloped past the crowd held its head high.4.82The pony was galloped past the crowd. 4.50

The virus mutated in the dish was extremely virulent. 4.33The virus that was mutated in the dish was extremely virulent.5.09The virus was mutated in the dish. 5.80

The plane coasted to a safe landing had a damaged engine.3.33The plane that was coasted to a safe landing had a damagedengine. 5.18The plane was coasted to a safe landing. 5.20

The teens trekked through Europe were unhappy. 3.10The teens that were trekked through Europe were unhappy.4.58The teens were trekked through Europe. 3.91

The soldiers marched for five hours straight were exhausted.2.70The soldiers that were marched for five hours straight wereexhausted. 5.67The soldiers were marched for five hours straight. 5.09

The prisoner snuck out of the jail at night was taken to a nearbywarehouse. 2.50The prisoner who was snuck out of the jail at night was taken toa nearby warehouse. 5.75The prisoner was snuck out of the jail at night. 5.18

The car slid across the icy road was actually part of a stunt for amovie. 2.60The car that was slid across the icy road was actually part of astunt for a movie. 4.67The car was slid across the icy road. 4.27

Appendix B

Baseline comparison sentences for hard and easy internal causereduced relatives

Anomalous sentences (found to be less acceptable than hard

internal cause reduced relatives)

The purple cow designed bushes math.Put the plate on the napkin on the bench yesterday.When the light turned green, everyone put and brought.The doctor said and stated for hours.The bartender grasped the mug of beer had been given to thewrong customer.That there exist politicians who like to talk and never seem topay attention to anyone else even though it is their job to is wellknown.Getting a job isn’t the easiest thing for a person to refrain fromdoing these days.

Comprehensible sentences (equal to easy internal cause reduced

relatives in acceptability)

That smoking is bad for you is a well known fact.The biotech industry banks on the Medicare law for help ondrug bills.A little before they thought that it would, spring arrived.The committee’s annoyance at the mayor’s insistence on beingheard at every meeting was getting obvious.John and Susan were having an affair after hours in the lab lastyear.


A sports announcer and her husband went to dinner at hermother’s house.Because my watch constantly loses time, I’m always late.The leaves were turning colors, and her thoughts turned towardwinter.In the hallway there ticked a grandfather clock.A convenience store was robbed by two teenagers but they gotaway with only $200.Only six of the smallest doughnuts were left.The bananas were way too ripe to eat.

Appendix C. Instructions for Study 2

After reading the first sentence below, please rate howacceptable it is as an English sentence, using a scale of 1–7. Arating of 1 indicates that the sentence is extremely unacceptable,while a rating of 7 would indicate the sentence is completelyacceptable. After you are done rating the sentence, shortenthe sentence’s length by deleting two or more words. You can-not add any new words or rearrange the words. Try not tochange the meaning of the sentence. Write out the shortenedsentence in the space provided, then rate how acceptable youfeel the shortened sentence is, also using a scale from 1 to 7.While we want you to try to produce good sentences with assimilar a meaning as you can, not all sentences can be shortenedin a grammatically correct fashion and you may sometimeshave trouble preserving the meaning of the first sentence. Yourlast task for this sentence is to decide how similar the meaningof the shortened sentence is to that of the original sentence,using a scale from 1 to 5. A rating of 1 would mean the sentenc-es have very different meanings, while a rating of 5 would meanthe sentences have the same meaning. After you are done withthe first sentence, go on to the rest in sequential order. If youhave any questions regarding the tasks, please let the experi-menter know now.

EXAMPLES:

1. The teenager was taping some music for his friend when hisstereo stopped working.
Acceptability rating: 1 2 3 4 5 6 7
1a. The teenager was taping music for his friend when his stereostopped.
Acceptability rating: 1 2 3 4 5 6 7 Similarity rating: 1 2 3 4 5
2. John was planning not to go on to the store.
2a. John was planning to go to the store.
Acceptability Rating: 1 2 3 4 5 6 7 Similarity rating: 1 2 3 4 5
3. Tarzan knifed the large leopard.
3a. Tarzan knifed leopard.
Acceptability rating: 1 2 3 4 5 6 7 Similarity rating: 1 2 3 4 5
References

Bever, T. (1970). The cognitive basis for linguistic structures. InJ. R. Hayes (Ed.), Cognition and the development of language

(pp. 279–362). New York: Wiley.Bolinger, D. L. (1975). On the passive in English. LACUS

Forum, 1, 57–80.Bolinger, D. L. (1977). Meaning and form. New York:

Longman.Carrow-Woolfolk, E. (1999). Comprehensive assessment of

spoken language. Circle Pines, NM: American GuidanceService.

Charniak, E. (1997). Statistical parsing with a context-freegrammar and word statistics. In Paper presented at the

AAAI-97, Providence, RI.Chomsky, N. (1957). Syntactic structures. The Hague: Mouton.Chomsky, N. (1965). Aspects of the theory of syntax. Cam-

bridge, MA: MIT Press.Clifton, C., Jr., Traxler, M. J., Mohamed, M. T., Williams,

R. S., Morris, R. K., & Rayner, K. (2003). The use ofthematic role information in parsing: Syntactic processingautonomy revisited. Journal of Memory and Language, 49,317–334.

Cohen, J. D., MacWhinney, B., Flatt, M., & Provost, J. (1993).PsyScope: A new graphic interactive environment fordesigning psychology experiments. Behavioral Research

Methods, Instruments & Computers, 25, 257–271.Ferreira, F., & Clifton, C. Jr., (1986). The independence of

syntactic processing. Journal of Memory and Language, 25,348–368.

Fisher, C., Gleitman, H., & Gleitman, L. R. (1991). On thesemantic content of subcategorization frames. Cognitive

Psychology, 23, 331–392.Frazier, L. (1987). Sentence processing: A tutorial review. In M.

Coltheart (Ed.), Attention and performance XII: The psy-

chology of reading (pp. 559–586). Hillsdale, NJ: Erlbaum.Frazier, L. (1995). Constraint satisfaction as a theory of

sentence processing. Journal of Psycholinguistic Research,

24, 437–468.Gillette, J., Gleitman, H., Gleitman, L., & Lederer, A. (1999).

Human simulations of vocabulary learning. Cognition, 73,135–176.

Gleitman, L. (1990). The structural source of verb meaning.Language Acquisition, 1, 3–55.

Goldberg, A. E. (1995). Constructions: A construction grammar

approach to argument structure. Chicago, IL: University ofChicago Press.

Goldberg, A. E. (2003). Constructions: A new theoreticalapproach to language. Trends in Cognitive Science, 5,219–224.

Goldberg, A. E., & Jackendoff, R. (2004). The Englishresultative as a family of constructions. Language, 80,532–568.

Hale, K., & Keyser, J. (1993). On argument structure and thelexical representation of syntactic relations. In K. Hale & J.Keyser (Eds.). The view from Building (Vol. 20, pp. 53–110).Cambridge, MA: MIT Press.

Hare, M., McRae, K., & Elman, J. L. (2004). Admitting thatadmitting verb sense into corpus analyses makes sense.Language and Cognitive Processes, 19, 181–224.


Hopper, P. J., & Thompson, S. (1980). Transitivity in grammarand discourse. Language, 56, 251–299.

Jackendoff, R. (1997). Twistin’ the night away. Language, 73,534–559.

Jackendoff, R. (2002a). English particle constructions, thelexicon, and the autonomy of syntax. In N. Dehe, R.Jackendoff, A. McIntyre, & S. Urban (Eds.), Verb-particle

explorations (pp. 67–94). Berlin: Mouton de Gruyter.Jackendoff, R. (2002b). Foundations of language. Oxford:

Oxford University Press.Lakoff, G. (1977). Linguistic gestalts. Proceedings of the

Chicago Linguistic Society, 13, 236–287.Levin, B. (1993). English verb classes and alternations: A

preliminary investigation. Chicago: University of ChicagoPress.

Levin, B., & Rappaport Hovav, M. (1995). Unaccusativity:

At the syntax-semantics interface. Cambridge, MA: MITPress.

MacDonald, M. C. (1993). The interaction of lexical andsyntactic ambiguity. Journal of Memory and Language, 32,692–715.

MacDonald, M. C., Pearlmutter, N. J., & Seidenberg, M. S.(1994). The lexical nature of syntactic ambiguity resolution.Psychological Review, 101, 676–703.

Marcus, M., Santorini, B., & Marcinkiewicz, M. A.(1993). Building a large annotated corpus of English:The Penn Treebank. Computational Linguistics, 19,313–330.

Masson, M. E. J., & Loftus, G. R. (2003). Using confidenceintervals for graphically based data interpretation. Canadian

Journal of Experimental Psychology, 57, 203–220.McKoon, G., & Macfarland, T. (2000). Externally and inter-

nally caused change of state verbs. Language, 76, 833–858.McKoon, G., & MacFarland, T. (2002). Event templates in the

lexical representations of verbs. Cognitive Psychology, 45,1–44.

McKoon, G., & Ratcliff, R. (2003). Meaning through syntax:Language comprehension and the reduced relative clauseconstruction. Psychological Review, 110, 490–525.

McKoon, G., & Ratcliff, R. (2005). ‘‘Meaning throughsyntax’’ in sentence production and comprehension: Replyto McRae et al. (2005). Psychological Review, 112,1032–1039.

McRae, K., Hare, M., & Tanenhaus, M. K. (2005). MeaningThrough Syntax is insufficient to explain comprehension ofsentences with reduced relative clauses: Comment onMcKoon and Ratcliff (2003). Psychological Review, 112,1022–1031.

McRae, K., Spivey-Knowlton, M. J., & Tanenhaus, M. K.(1998). Modeling the influence of thematic fit (and otherconstraints) in on-line sentence comprehension. Journal of

Memory and Language, 38, 283–312.Merlo, P., & Stevenson, S. (1998). What grammars tell us about

corpora: the case of reduced relative clauses. In Proceedings

of the 6th workshop on very large corpora, Montreal,Canada, pp. 134–142.

Pearlmutter, N., & MacDonald, M. (1995). Probabilisticconstraints and working memory capacity in syntacticambiguity resolution. Journal of Memory and Language,

43, 521–542.

Pinker, S. (1989). Learnability and cognition. Cambridge, MA:MIT Press.

Pollatsek, A., & Well, A. D. (1995). On the use ofcounterbalanced designs in cognitive research: A sugges-tion for a better and more powerful analysis. Journal of

Experimental Psychology: Learning, Memory, & Cognition,

21, 785–794.Pustejovsky, J. (1995). The generative lexicon. Computational

Linguistics, 17, 109–441.Rayner, K., Carlson, M., & Frazier, L. (1983). The interac-

tion of syntax and semantics during sentence processing.Journal of Verbal Learning and Verbal Behavior, 22,358–374.

Rayner, K., Garrod, S., & Perfetti, C. (1992). Discourseinfluences in parsing are delayed. Cognition, 45, 109–139.

Rice, S. (1987). Towards a transitive prototype: Evidence fromsome atypical English passives. Proceedings of the Berkeley

Linguistic Society, 13, 422–434.Roland, D., Dick, F., & Elman, J. Frequency of basic English

grammatical structures: A corpus analysis. Submitted forpublication.

Roland, D., Elman, J. L., & Ferreira, V. (2006). Why isthat? Structural prediction and ambiguity resolution in avery large corpus of English sentences. Cognition, 98,245–272.

Spivey-Knowlton, M., Trueswell, J. C., & Tanenhaus, M. K.(1993). Context effects in syntactic ambiguity resolution:Discourse and semantic influences in parsing reducedrelative clauses. Canadian Journal of Experimental Psychol-

ogy, 37, 276–309.Stevenson, S. (1994). Competition and recency in a hybrid

network model of syntactic disambiguation. Journal of

Psycholinguistic Research, 23, 295–322.Stevenson, S., & Merlo, P. (1997). Lexical structure and

parsing complexity. Language and Cognitive Processes, 12,349–399.

Sturt, P., Scheepers, C., & Pickering, M. J. (2002).Syntactic ambiguity resolution after initial misanalysis:The role of recency. Journal of Memory and Language,

46, 371–390.Tanenhaus, M. K., & Grodner, D. (2006). Sentence compre-

hension in adults: Behavioral studies. In M. J. Spivey, M.Joanisse, & K. McRae (Eds.), Cambridge handbook of

psycholinguistics. Cambridge: Cambridge University Press,in press.

Tanenhaus, M. K., Spivey-Knowlton, M. J., & Hanna, J. E.(2000). Modeling thematic and discourse context effects onsyntactic ambiguity resolution within a multiple constraintsframework: Implications for the architecture of the lan-guage processing system. In M. Pickering, C. Clifton, & M.Crocker (Eds.), Architecture and mechanisms of the language

processing system (pp. 90–118). Cambridge, MA: Cam-bridge University Press.

Tanenhaus, M. K., & Trueswell, J. C. (1995). Sentence compre-hension. In Handbook of perception and cognition. In J. Miller& P. Eimas (Eds.). Speech, language and communication (Vol.11, pp. 217–262). New York: Academic Press.

Tomasello, M., & Stahl, D. (2004). Sampling children’sspontaneous speech: how much is enough? Journal of Child

Language, 31, 101–121.


Trueswell, J. C. (1996). The role of lexical frequency in syntacticambiguity resolution. Journal of Memory and Language, 35,566–585.

Trueswell, J. C., & Tanenhaus, M. K. (1991). Tense, temporalcontext and syntactic ambiguity resolution. Language and

Cognitive Processes, 6, 303–338.

Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S.(1994). Semantic influences on parsing: Use ofthematic role information in syntactic ambiguity res-olution. Journal of Memory and Language, 33,285–318.

Understanding and producing the reduced...

Documents

Transcript of Understanding and producing the reduced...